Review

A Survey of Forex and Stock Price Prediction Using Deep

Learning
Zexin Hu1^ , Yiqi Zhao2^ and Matloob Khushi3,*

^ authors contributed equally both authors should be considered first-author

1 Affiliation 1; [email protected]
2 Affiliation 2; [email protected]
2 Affiliation 3; [email protected]

* Correspondence [email protected];

Abstract: The prediction of stock and foreign exchange (Forex) had always been a hot and profitable
area of study. Deep learning application had proven to yields better accuracy and return in the field
of financial prediction and forecasting. In this survey we selected papers from the DBLP database
for comparison and analysis. We classified papers according to different deep learning methods,
which included: Convolutional neural network (CNN), Long Short-Term Memory (LSTM), Deep
neural network (DNN), Recurrent Neural Network (RNN), Reinforcement Learning, and other deep
learning methods such as HAN, NLP, and Wavenet. Furthermore, this paper reviewed the dataset,
variable, model, and results of each article. The survey presented the results through the most used
performance metrics: RMSE, MAPE, MAE, MSE, accuracy, Sharpe ratio, and return rate. We identi-
fied that recent models that combined LSTM with other methods, for example, DNN, are widely
researched. Reinforcement learning and other deep learning method yielded great returns and per-
formances. We conclude that in recent years the trend of using deep-learning based method for
financial modeling is exponentially rising.

Keywords: Deep Learning; Stock; Foreign Exchange; Financial Prediction; Survey

Citation: Lastname, F.; Lastname, F.;

Last-name, F. Title. Appl. Syst. Innov.
1. Introduction
2021, 4, x.
https://doi.org/10.3390/xxxxx
The share market is a snapshot of future growth expectation of the companies as well
as the economy. Many factors have attributed to the stock's price fluctuation, which in-
Received: date cludes but not limited to macro-economic factors, the market anticipation and confidence
Accepted: date in the company's management and operation. The advancement of technology allows the
Published: date public to access a larger quantity of information in a timelier manner. It means that stock
analysis has become more and more difficult as a considerable amount of data has to be
Publisher's Note: MDPI stays neu- processed in a relatively short time. People hope that the progress made in big data, es-
tral with regard to jurisdictional pecially in the Deep Learning field, can help them analyses stock information [1].
claims in published maps and insti- Forex is one of the largest financial markets in the world. The prediction of the ex-
tutional affiliations. change rate can provide investors with useful decision-making references to increase re-
turn and reduce risk. However, the exchange rate is always under the influence of many
factors, such as countries' economy, politics, society, international situation, etc., so the
complexity of the matter has made Forex prediction and forecasting a challenging re-
Copyright: © 2020 by the authors.
Submitted for possible open access
search topic.[2] Nowadays, Forex Forecasting tasks apply many different deep learning
publication under the terms and
models as the computer, and artificial intelligence technology matures.
conditions of the Creative Commons Forex and stock are similar in many aspects. For example, they both have comparable
Attribution (CC BY) license technical indicators, both have similar charts (candle chart), and they would both be af-
(http://creativecommons.org/licenses fected by its country's market sentiment. Therefore, this paper will discuss the application
/by/4.0/).

Appl. Syst. Innov. 2021, 4, x. https://doi.org/10.3390/xxxxx www.mdpi.com/journal/asi

Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 2 of 34

of deep learning to both Forex and the stock market and explore the impact of different
deep learning methods on their price trend prediction accuracy.
The continuous development in the AI field leads to the wide use of deep learning
techniques in many research fields and practical scenarios. Applications include natural
language processing, image recognition, medical predictions, and more. The neural net-
works used in these applications have also developed and improved due to the rise of
deep learning. For example, reinforcement learning has gained popularity since AlphaGo
defeated the best chess player at the time using it, and reinforcement learning has been
implemented in the financial prediction field since then [76]. These technological break-
throughs have given the stock and Forex prediction models a solid foundation to start and
a greater room to improve.
The highly complex nonlinear relationship of deep learning can fully describe the
complex characteristics of the influencing factors. Besides, many other fields have verified
the accuracy of a deep learning model for prediction accuracy, such as image classifica-
tion, gene analysis. Research results are also obtained for Time-series data analysis and
prediction with a deep learning algorithm; for example, deep learning is used to predict
the offline store traffic [3]. Overall, deep learning models have excellent performance in
other research fields. Therefore, it is feasible to predict stock and Forex trends with deep
learning.
Financial researchers around the world have been studying and analysing the
changes in the stock and Forex market. The broadening application of artificial intelli-
gence has led to an increasing number of investors using deep learning model to predict
and study the stock and Forex price. It has been proven that the fluctuation of stock and
Forex price could be predicted [4]. Different from the traditional statistical and economet-
ric models, deep learning can describe complex influencing factors.
Therefore, this paper will investigate the different effects of different deep learning
methods on stock and Forex forecasting according to the existing published papers. This
survey will analyse each paper from the following aspects: 1. What is the dataset of this
paper; 2.What is the variable of this paper;3. What kind of deep learning model had been
adopted; 4. What is the result of the prediction model?
The structure of this paper will be as follows: firstly, the introduction of Forex and
stock combined with deep learning; secondly, the criteria and research methods of the
article selected by the survey, thirdly, the impact and analysis of different deep learning
methods on stocks and Forex prediction; fourthly, the discussion and analysis of the above
methods; finally, the conclusion of the whole paper.

2 Related deep learning methods and input introduction

2.1 Convolutional neural network (CNN)
CNN was widely used in the field of image recognition because of its powerful pat-
tern recognition ability; its use was also extended to the field of economic prediction. Sim-
ilar to the traditional neural network, CNN was composed of multiple neurons connected
by a hierarchical structure, and the weights and bias between layers can be trained. CNN
was different from the network structure of Fully Connected network such as
DBN/SAE/BP, as the CNN could share the weight among the neurons in each layer of the
network. Hence the model significantly reduced the weight of the network and avoided
falling into dimensional disaster and local minimisation [5].
If the characteristics of the stock market at a specific time point were regarded as a
feature graph, CNN had the potential to extract the characteristics of the stock market at
the corresponding period from these feature graphs. Therefore, CNN could be used to
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 3 of 34

build a timing-selection model and ultimately used to complete the construction of the
timing-selection strategy.

2.2 Recurrent Neural Network (RNN)

RNN belonged to the neural network, and it was good at modelling and processing
sequential data. The specific expression was that the RNN was able to memorise the pre-
vious state, and the previous state could be used in the current state calculation. The dif-
ferent hidden layers were non-independent, and the input of the current hidden layer
included not only the output of the input layer but also the output of the previously hid-
den layer. For this reason, RNN would have a good performance in dealing with the se-
quential data.
The advantage of RNN was that it would consider the context of data in the process
of training, which is very suitable for the scenario of stocks and Forex because the fluctu-
ation at a particular time often contains some connection with the previous trend.
2.1.3 Long Short-Term Memory (LSTM)
LSTM model is one of the variants of the RNN. Its core contribution was to introduce
the design of self-loop to generate the path of the gradient, which could continuously flow
for an extended period. The weight of the self-loop was also updated in each iteration,
which solved the gradient vanishing problem that was easily generated when the RNN
model updated the weights [6].
The modelling of time series was essentially a process of nonlinear parameter fitting.
The LSTM model could perform well to reveal the correlation of nonlinear time series in
the delay state space, and to realise the purpose of stock prediction [7]. In the stock or
Forex trend prediction model based on LSTM, it obtained the corresponding data charac-
teristics from the stock or Forex history data.

2.4 Deep neural network (DNN)

DNN was a neural network with at least one hidden layer. It could provide model-
ling for complicated nonlinear functions and own a high-level abstraction ability which
meant the fitting power of the model would be significantly improved. Meanwhile, it was
a kind of discriminant model, which could be trained through the backpropagation algo-
rithm.
Since the DNN was good at dealing with prediction problems with sizable data and
complicated nonlinear mapping relations, an intelligent stock and Forex prediction sys-
tem can be designed based on a DNN to predict stock and Forex trend. Hopefully, the
model was able to achieve far higher accuracy than human beings.

2.5 Reinforcement learning

Reinforcement learning was one of the deep learning methods that focus on how to
act according to the current situation to profit maximisation. In reinforcement learning,
there were two basic elements: state and action. A strategy was defined as performing a
particular action in a specific state. All the learner had to do was to learn a good strategy
by continually exploring and learning.
If the state is regarded as the attribute and the action as the label, it was easy to know
that both supervised learning and reinforcement learning was trying to find a map and
inferred the label/action from the known attribute/state. In this way, the strategy in rein-
forcement learning was equivalent to the classification/regression in supervised learning.
However, in practical problems, reinforcement learning did not have such labelling infor-
mation as supervised learning, and the results were often obtained after the attempt of the
action. Therefore, reinforcement learning would continuously adjust the previous
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 4 of 34

strategy through the feedback of the result information, for this reason, the algorithm
would learn: in which state to take which step would have the most beneficial result.
Therefore, reinforcement learning could learn the best timing trading action (select-
ing the best price, trading duration, and order size) according to the market response. It
can view the contextual information (price, news, public opinion, interest rate, fees, trad-
ing action, returns, losses, etc.) of the transaction as an environment of reinforcement
learning. Gains or losses could be thought as the reward for learning, trading actions as
actions, and factors as states or observations to realise the prediction of the stock and Forex
trend.

2.5 Other deep learning methods

In this paper, we will discuss the prediction of stock and Forex trend by other deep
learning methods, for example, Hybrid Attention Networks (HAN) and self-paced learn-
ing mechanism (NLP); multi-filters neural network (MFNN), and Wavenet. The frequency
of these methods in the selected articles is not high so that they will be discussed in section
4.7.

3. Review methodology and criteria

3.1 Paper selection methods
In the past few years, there had been many deep learning model papers on stock and
Forex forecasting. In this paper, the articles analysed in this paper were all from the DBLP
computer science bibliography and Microsoft Academic.
Firstly, the keywords were searched in DBLP are:' CNN stock/Forex';' LSTM
stock/Forex';' Deep learning stock/ Forex';' RNN stock/Forex' and' Reinforcement learning
stock/Forex'. The keywords were searched in Microsoft Academic, then the filter of ‘2015-
2021’ and ‘Deep learning’ were applied.
Secondly, The quality of the selected articles from DBLP was ensured by excluding
all the journals and conferences that were informally published. The quality of the selected
articles from Microsoft Academic were controlled by excluding all the journals and con-
ferences that were informally published as well as a minimum of 5 citations. Filtering on
citations as the searching method was designed to detect under-study area with in this
field, the paper with 5 more citations can be an indication that the area could be potentially
explored. Furthermore, it should be noted that we only consider papers with a novel
model, implementing existing model would not be analysised in this review.
At the same time, the timeliness of the survey was provided by focusing on publica-
tions after 2015. Among them, there were 4 papers in 2015,1 papers in 2016, 15 papers in
2017, 30 papers in 2018, 28 papers in 2019, and 10 papers in 2020, in total 88 existing papers
would be reviewed.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 5 of 34

Figure 1 The systematic literature review in this paper

3.2 Selected paper statistics

In this paper, the goal was to study the use of six different deep learning methods in
Forex/stock (which are CNN, LSTM, DNN, RNN, Reinforcement learning, and other deep
learning methods) with different datasets, input variables, and model type. All the results
were compared together for discussion and for concluding after individually analysed.
Figure 1 showed the annual distribution of papers collected and reviewed. Figure 2
showed the distribution of the different methods in this paper. Figure 3 showed the sys-
tematic literature review in this paper. Table 1 showed the distribution of different types
of articles in this paper.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 6 of 34

2020

2019

2018

2017

2016

2015

0 5 10 15 20 25 30 35

CNN LSTM DNN RNN Reinforcement learning Other deep learning methods

2020

2019

2018

2017

2016

2015

0 5 10 15 20 25 30

CNN LSTM DNN RNN Reinforcement learning Other deep learning methods

Figure 2 The annual distribution of papers collected and reviewed

Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 7 of 34

CNN LSTM DNN RNN Reinforcement learning Other deep learning methods

11%
21%
9%

6%

10%

43%

CNN LSTM DNN RNN Reinforcement learning Other deep learning methods

12%
19%

12%

7%

10%
40%

Figure 3 The distribution of the different methods in this paper

Table1. The distribution of different types of articles in this paper

Type of Method Total paper Number of papers in journals Number of papers in conferences
CNN 18 6 12
LSTM 38 13 25
DNN 9 5 4
RNN 5 2 3
Reinforcement Learning 8 1 7
Other deep learning methods 10 3 7

4. Results
4.1 Papers Descriptions
4.1.1 CNN
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 8 of 34

17 articles that used CNN technology for the stock and Forex prediction were briefly
described below, and as shown in the table. Table 2 shows the author, variables, dataset,
and model of all the papers mentioned.
Table 2 The overall information for the papers which used the CNN model
Reference No. Author Dataset Variables Model
[8] Maqsood, H. 1.Top 4 perform- 1.Open price, CNN
ing companies of high price, low price,
US, Hong AdjClose price, Volume,
Kong, Turkey, and and Close price
Pakistan 2.sentiment(Positive, neu-
2.Twitter dataset tral and negative senti-
ment).

[9] Patil, P. 1. News collected Adjacency matrix calcu- Graph convolu-

on a financial lated by the correlation tion neural net-
website. coefficient and using news work (GCN)
2.30 stocks of for- co-mentions
tune 500 compa-
nies, such as
WMT, XOM, and
AAPL
[10] Sim, HS. S&P 500 minute Close price CNN
data
[11] Hoseinzade, E. S & P 500, 82 variables including high, 2D-CNN
NASDAQ, Dow low, close price; Volume, And 3D-CNN
Jones industrial RSI, KD, WR.etc
average, NYSE,
and Russell
[12] Eapen, J. Standard and Close price CNN--Bi-Direc-
Poor's (S&P) 500 tional
stock dataset LSTM
[13] Yang, H. S&P 500 Index high, low,close price;Vol- CNN with MICFS
ETF (SPY) ume,RSI,KD,WR,ROC,CCI

[14] Cai, S. 1.Crawling finan- 1.word vector; headline and CNN-LSTM

cial news keyword training set in the
2.Baidu Index news
2.Close price
[15] Oncharoen, P. Reuters and Red- Word vectors of headlines CNN-LSTM
dit Close prices, Bollinger
Standard & Poor's band, RSI, and Stochastic
500 Index Oscillator
(S&P500) and
Dow Jones Indus-
trial Average
(DJIA)
[16] Liu, Y. 1.Thomson Reu- 1.Financial news corpus TransE-CNN-
ters with headlines from Apple LSTM
2.Standard & 2.. Open price; Close
Poor’s 500 index price; High price; Low
(S&P 500) price; Volume; Stochastic
Oscillator (%K); Larry Wil-
liam (LW) %R indicator;
and Relative Strength
Index (RSI)
[17] Selvin, S. NSE listed com- Close price CNN Sliding-win-
panies dow model
: Infosys, TCS,
and CIPLA
[18] Liu, S. Chinese stocks Close price CNN-LSTM
from the SINA FI-
NANCE(Not
given specific
data)
[19] Gudelek, M.U. Exchange- Close price, RSI, SMA, 2D-CNN
Traded MACD, MFI, Wil-
Funds(ETF) liams %R, the stochastic os-
cillator, the ultimate oscilla-
tor
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 9 of 34

[20] Ding, X. 1.S&P 500 index 1.close price NTN-CNN

2.Reuters and 2.long-term, mid-term and
Bloomberg short-term feature vectors
of news headlines
[71] Zhao, Y. and Khu- USDJPY Ex- 1. Momentum Indicators: Wavelet De-
shi change rate RSI5, RSI10, RSI20, noised-ResNet
MACD, MACDhist, MAC- CNN with light
Dsignal, Slowk, Slowd, GBM
Fastk, Fastd, WR5, WR10,
WR20, ROC5, ROC10,
ROC20, CCI5, CCI10,
CCI20
2. Volume Indicators:
ATR5, ATR10, ATR20 ,
NATR5, NATR10,
NATR20, TRANGE
[76] Chen, S. and He, Chinese stock Closing price CNN
H. market
[77] Ugur Gudelek, M. SPDR ETF Closing price and technical CNN
indicators
[78] Chen, Jou-Fan Taiwan index fu- Open, high, low, closing GAF+CNN
tures price
[90] Wen, M S%P500 Open, High, Low, Close, CNN
Adjusted Close, and Vol-
ume.

Maqsood, H. proposed a CNN model that made use of the price and sentiment anal-
ysis as input and compared the proposed model with the Linear Regression and SVM. He
concluded that not all the significant events have a serious impact on stock exchange pre-
diction. However, more important local events could affect the performance of prediction
algorithms [8].
Patil, P. proposed a new network using graph theory and CNN, which leveraged
Spatio-temporal relationship information between different stocks by modelling the stock
market as a complex network. Meanwhile, the model used both stock indicators and fi-
nancial news as input [9].
Sim, HS proposed a CNN network that using the 9 technical indicators (Close Price,
SMA, EMA, ROC, MACD, Fast%K, Slow%D, Upper Band, Lower Band) to verify the ap-
plicability of the CNN method in the stock market. And he concluded that the use of tech-
nical indicators in stock price forecasting by CNN has no positive effect [10].
Hoseinzade, E. proposed two models: 2D-CNN and 3D-CNN using 82 different tech-
nical indicators. And These two structures could improve the predictive performance of
the baseline algorithm by about 3% to 11% [11].
Eapen, J.proposed a model that had multiple pipelines of CNN and bi-directional
LSTM units. And it could improve prediction performance by 9% using a single pipeline
deep learning model and by over a factor of six using support vector machine regressor
model on S&P 500 grand challenge dataset [12].
Yang, H.m, proposed a multi-indicator feature selection for stock index prediction
based on a multichannel CNN structure without sub-sampling, named MI-CNN frame-
work. In this method, candidate indicators were selected by the maximal information co-
efficient feature selection (MICFS) approach, to ensure the correlation with stock move-
ments whilst reducing redundancy between different indicators [13].
Cai, S. proposed the CNN and LSTM forecasting system with financial news and
historical data of the stock market. It had generated seven prediction models. According
to the ensemble learning method, the seven models are constructed into one ensemble
model to obtain an aggregated model. Unfortunately, all models had a lower prediction
accuracy [14].
Oncharoen, P. proposed a new framework to train a DNN for stock market predic-
tion. A new loss function was developed by adding a risk-reward function, which was
derived from the trading simulation results [15].
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 10 of 34

Liu, Y. proposed to incorporate a joint model using the TransE model for representa-
tion learning and a CNN, which extracted features from financial news articles. And then
he combined the model with LSTM to predict the price movement. The model could im-
prove the accuracy of text feature extraction while reducing the sparseness of news head-
lines [16].
Selvin, S. used CNN, LSTM, and RNN architectures to forecast the price of NSE listed
companies and compares their performance. The final results showed that CNN is the best
architecture to predict the price of the stock because it could identify the trend of the di-
rectional change [17].
Liu, S.proposed a CNN-LSTM model, and the model performed basic Momentum
strategy and Benchmark model whose return rates were 0.882 and 1.136 respectively. The
CNN part could extract useful features even from low signal-to-noise time-series data,
and the LSTM part could predict future stock prices with high accuracy, then the predict-
ing outcomes were used as timing signals [18].
Gudelek, M.U. proposed a 2D-CNN model. This model adopted a sliding window
approach then generated images by taking snapshots that are bounded by the window
over a daily period. The model could predict the next day's prices with 72% accuracy and
end up with 5:1 of our initial capital [19].
Ding, X. proposed a deep learning method for the event-driven stock market predic-
tion. Firstly, events were extracted from news text and represented as dense vectors,
trained using a novel neural tensor Network(NTN). Secondly, CNN was used to model
both short-term and long-term influences of events on stock price movements [20].
Zhao, Y. proposed a Wavelet Denoised-ResNet CNN for Forex exchange rate predic-
tion. The technical indicators were treated as an image matrix, the image matrix was first
denoised using Wavelet method, then processed by ResNet CNN, and lastly, LightGBM
was used to replace the softmax layer to output a prediction. [71]
Chen, S. and He, H. proposed a CNN model with a novel architecture, where it gen-
erated better results than benchmark RNN model. [76]
Ugur Gudelek, M proposed a CNN model where financial data is taken as a image,
then using CNN to predict the desired trading action, the model yielded reasonable re-
sults. [77]
Chen, Jou-Fan proposed a novel CNN model which utilized the power of Gramian
Angular Field, the results produced were average, but it is a interesting research direction.
[78]
Wen, M proposed a CNN model which relied on the reconstructed of time series, it
turned the time series into segment, then CNN was used to classify each segment. The
model generated good results.[90]
4.1.2 RNN
5 articles that used RNN technology for the stock and Forex prediction were briefly
described below, and as shown in the table. Table 3 showed the author, variables, dataset,
and model of all the papers mentioned above.
Table 3 The overall information for the papers which used RNN model
Reference no. Author Dataset Variables Model
[53] Ni, L. EURUSD,AU- open price, close C-RNN
DUSD,XAU- price,
USD,GBPJPY,EUR- highest price and
JPY,GBPUSD,USD lowest price
CHF,USDJPY and
USDCAD
[54] Li, C. CSI300,CSI200and Open price, high Multi-task RNN with
CSI500 price, low price, MRFs
close price, Volume,
and amount.
[55] Chen, W HS300 1.Technical features: RNN-Boost
open price, high
price, low price,
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 11 of 34

close price, Volume,

price change, price
limit, Volume
change, Volume
limit, Amplitude, and
difference.
2.Content features:
sentiment features
and LDA features
[56] Zhang, R. Sandstorm sector of Open price, close C-RNN(DWNN)
Shanghai Stock Ex- price, highest price,
change lowest price, and the
daily trading volume
[73] Zeng, Z. and Khushi USDJPY Exchange Momentum indica- Attention-based
rate tors: average direc- RNN-ARIMA
tional movement in-
dex, absolute price
oscillator, arron os-
cillator, balance of
power, commodity
channel index,
chande momentum
oscillator, percentage
price oscillator, mov-
ing average conver-
gence divergence,
Williams, momen-
tum, relative strength
index, stochastic os-
cillator, triple expo-
nential average. •
Volatility indicators:
average true range,
normalised average
true range, true
range.

Ni, L. proposed a CRNN model to predict the price of 9 kinds of forex pair. The results showed
that the proposed model performed much better than the LSTM model and the CNN model[53].
Li, C. proposed a Multi-task RNN model with MRFs. The MMPL is proposed to automati-
cally extract diversified and complementary features from individual stock price sequences which
means there is no need for the technical indicators. Features learned by MMPL were passed to a
binary MRF with a weighted lower linear envelope energy function to utilise intra-clique higher-
order consistency between stocks[54].
Chen, W proposed an RNN-Boost model that made use of the technical indicators and senti-
ment features, and Latent Dirichlet allocation (LDA) features to predict the price of stocks. Its
results showed that the proposed model outperformed the single-RNN model[55].
Zhang, R. proposed a Deep and Wide Neural Networks (DWNN) model where CNN's con-
volution layer was added to the RNN's hidden state transfer process. The results showed that the
DWNN model could reduce the prediction mean squared error by 30% compared with the general
RNN model[56].
Zeng. Z proposed a novel Attention-based RNN (ARNN) where wavelet denoised input was
passed to ARNN. The forecast was calculated using the ARIMA and the output of ARNN model.
[73]
4.1.3 LSTM
27 articles that used LSTM technology for the stock and Forex prediction were briefly
described below, and as shown in the table. Table 4 showed the author, variables, dataset,
and model of all the papers mentioned.
Table 4 The overall information for the papers which used LSTM model
# Author Dataset Variables Model
[21] Nikou, M. iShares MSCI United Close price LSTM
Kingdom
[22] Fazeli, A. S&P 500 Open price, high LSTM
price, low price,
close price, adjusted
close price, Volume,
volatility,
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 12 of 34

Williams %R and
RSI
[23] Xu, Y. Microsoft (MSFT)、 Open price, high Attention-based
PO logistics (XPO) price, low price, LSTM
and AMD Daily close
stock price data are price,AD,ADX,EMA
collected from Yahoo ,KAMA,MA,MACD,
Finance from 11 in- RSI,SAR,AMA .etc
dustries and finance tweet
Finance tweets from sentiment
a social media com-
pany StockTwits
[24] Lakshminarayanan, Dow Jones Industrial Close price, moving LSTM
S.K. Average (DJIA) average, Crud oil
price, gold price
[25] Rana, M. Spanish stock com- Close price LSTM
pany Acciona
[26] Naik, N. CIPLA stock, ITC Close price RNN with LSTM
stock,
TCS stock, ONGC
stock, and Nifty in-
dex
[27] Nguyen, D.H.D NASDAQ stock mar- Trade volume, open, Dynamic LSTM
ket:GE,AAPL,SNP close, high, low, and
and FB adjusted close prices
[28] Lai, C.Y. Foxconn, Quanta, KD, OBV, MACD, LSTM
Formosa RSI, and the average
Taiwan Cement and of the previous five
Taiwan Semiconduc- days stock market in-
tor formation (open,
high, low, Volume,
close)
[29] Hossain, M.A. S&P500 Open price, close LSTM with GRU
price,
volume
[30] Baek, Y. KOSPI200 and Close price LSTM with preven-
S&P500 tion module, predic-
tion module
[31] Kim, H.Y. KOSPI 200 Close price GEW-LSTM
[32] Li, H. CSI-300 Open price Attention-based
Multi-Input LSTM
[33] Cheng, L.-C. Data from Taiwan Open price, close Attention-based
Stock Exchange Cor- price, low price, high LSTM
poration((Not given price, Volume, KD,
specific data) MA, RSV, etc.
[34] Shah, D. Tech Mahindra Close price LSTM
(NSE: TECHM)
BSESensex
[35] Lin, B.-S. Taiwan Stock Ex- Trade volume, trans- LSTM
change Corporation action, open price,
(TWSE) highest price, lowest
price, close price,
KD, RSI, and Bol-
linger
Bands(BBands)
[36] Skehin, T. Facebook Inc. (FB), Close price ARIMA-LSTM-
Apple Inc. (AAPL), Wavelet
Amazon.com Inc
(AMZN), Netflix
Inc.
(NFLX) and Alpha-
bet Inc. (GOOG) in
NASDAQ of S&P
500
[37] Zhang, X. China's A-share mar- The open price, close RNN with LSTM
ket price, highest price,
lowest price, and
trading volume and
11 indicators
[38] Achkar, R. Facebook Close price RNN with LSTM
stocks, Google
stocks, and Bitcoin
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 13 of 34

stocks collected from

Yahoo finance
[39] Zeng, Y. SSE50 index N/A LSTM
[40] Shao, X. Ping An Bank Close price Kmeans-LSTM
[41] Zhao, Z. SSE Composite In- Close price Time-weighted
dex,SSE 50,CSI LSTM
500,CSI 300 and
SZSE Composite In-
dex
[42] Nelson, D.M. IBovespa index from Open price, close LSTM
the BM&F Bovespa price, high price, low
stock exchange price and Volume ex-
ponentially weighted
moving averages,
etc.(175 indicators in
total)
[43] dos Santos Pinheiro 1.Standard&Poor's 1.financial domain NLP+LSTM
500 index news text(headlines
2.Reuters and instead of the full
Bloomberg content); 2.Close
price
[44] Buczkowski, P. Expert recommenda- A stock identifier, a LSTM with GRU
tion from TipRanks list of expert recom-
company mendations of vary-
ing length, and op-
tional target labels
(class)
[3] Akita, R. Morning edition of 1.Paragraph Vector; LSTM
the Nikkei newspa- 2open, close, highest
per and lowest price
Nikkei 225
[45] Chen, K China stock market volume, high, low, LSTM
in Shanghai and open, close price
Shenzhen from Ya-
hoo finance
[75] Qi, Ling and Khushi Forex Exchange rate Technical indicators LSTM
[80] Pang, Xiong Wen Chinese stock stock data, stock DeepLSTM with en-
news, capital stock coder
and shareholders, and
financial analysis
[81] Feng, Fuli, et al. NASDAQ and Stock data LSTM with ranking
NYSE relation
[82] Chung, Hyejung, and Korea Stock Price In- Open, High, low, LSTM with GA
Kyung-shik Shin. dex closing, volume and
technical indicators.
[83] Li, Jiahong, et al. Chinese Stock Mar- Closing price and LSTM with Naïve
ket sentiment data Bayes
[84] Zhang, Kang, et al. S&P 500 Index, Open, High, low, LSTM with Genera-
ShanghaiComposite closing, volume and tive Adversarial Net-
Index, IBM from technical indicators. work (GAN)
NYSE, MSFT from
NASDAQ, PingAn
Insurance Company
of China (PAICC)
[91] Jin, Zhigang, et al. Apple stock price Sentiment and stock LSTM with senti-
price data ment analysis model
[92] Long, Jiawei, et al. Chinese Stock mar- Market information LSTM with CNN
ket including transaction
records
[93] Chen,M. Chinese stock market Sentiment infor- LSTM
mation
[94] Qian,F. Chen,X Chinese stock market Closing price LSTM with ARIMA
[95] Li,Z. Tam,V. Asian stock indexes Closing price and LSTM with Wavelet
technical indicators Denosing

Nikou, M. proposed an LSTM model and compared it with the ANN model, SVR
model, and RF model. The results showed that the LSTM model performed better in the
prediction of the close price of iShares MSCI United Kingdom than the other models men-
tioned in the paper[21].
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 14 of 34

Fazeli, A. proposed an LSTM model; his result showed an improve through using
LeakyReLU. MSE came below 0.1. Whilst compared with the ReLU function; the MSE
went well above 0.2. He examined the effect of RSI, Williams %R, and volatility on the loss
of the model. It was shown that the model's loss was reduced by using only RSI, which
contributed to the performance of the model[22].
Xu, Y. proposed an attention-based LSTM model that performs better than the regu-
lar LSTM. It was found that finance tweets that were posted from market closure to market
open the next day had more predictive power on the next day's stock movement. The
weighted sentiment on the max follower on StockTwits also performed much better than
other methods[23].
Lakshminarayanan, S.K. proposed an LSTM model combined with Crud oil price,
gold price, and moving average, which performed much better than the LSTM model
without them and the SVM model. It showed that the Crud oil and gold price had some
impact on the stock price prediction[24].
Rana, M. proposed an LSTM model that outperformed the LR and SVR model. He
also compared the different activation functions with different optimisers and concluded
that the tanh activation with the adam algorithm performs best with the accuracy of
98.49%[25].
Naik, N. proposed an RNN with the LSTM model. The model had the ability to keep
the memory of historical stock returns in order to forecast future stock return output. It
was worth noting that the recent stock information rather than old related stock infor-
mation was stored and used. The network also outperformed the Feed Forward ANN
model[26].
Lai, C.Y. proposed a Dynamic LSTM model The results showed that stock prediction
accuracy based on MAE, MAPE, RMSE, and MSE obtained by the dynamic LSTM model
was much better than that by the static LSTM model. The dynamic model also consistently
outperformed the linear models SA-5 and EMA-0.5 when predicting four stocks[27].
Lai, C.Y. proposed an LSTM model which used average previous five days stock
market information (open, high, low, Volume, close) as the input value. The initial predic-
tion was calculated using the value. The prediction was then used as part of the average
of the stock price information for the next five days through the ARIMA method. Moreo-
ver, he utilised Technical Analysis Indicators to consider whether to buy stocks or con-
tinue to hold stocks or sell stocks[28].
Hossain, M.A. proposed an LSTM model followed by GRU. Both of the LSTM and
GRU were powerful recurrent networks that could perform better and faster in terms of
accuracy in regression-based prediction. And the proposed model outperformed the
LSTM only, GRU only, and GRU followed by the LSTM model[29].
Baek, Y. proposed a novel data augmentation approach for stock market index fore-
casting through ModAugNet framework, which consisted of two modules: an overfitting
prevention LSTM module and a prediction LSTM module. The overfitting problems
mainly caused by the limited availability of data points for training[30].
Kim, H.Y. proposed some LSTM model to forecast stock price volatility that com-
bined the LSTM model with various generalised autoregressive conditional heteroscedas-
ticity (GARCH) -type models. He found that the GEW-LSTM model, which combined all
three models, GARCH, EGARCH, and EWMA, with LSTM, performed best[31].
Li, H. proposed an improved MI- LSTM based on LSTM and attention mechanism,
which achieved better performance in extracting potential information and filtering noise.
The model could assign different weights to different input series to keep the dominant
status of the mainstream while absorbing information from leaky input gates[32].
Cheng, L.-C. proposed an Attention-based LSTM model that could solve the problem
of exploding and vanishing gradients and thus did not effectively capture long-term de-
pendencies[33].
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 15 of 34

Shah, D. proposed an LSTM model which was compared with the DNN model. The
proposed model was able to predict volatile movements in the true data. In general, it was
able to recognise the directionality of the changes in the true data more accurately than
the DNN[34].
Lin, B.-S. proposed an LSTM model to predict the price of the top 10 industries in-
cluded in Taiwan Stock Exchange Corporation (TWSE).In the experimental results, most
of the results were reasonable except for the MTK stock[35].
Skehin, T. proposed a linear Autoregressive Integrated Moving Average (ARIMA)
model and LSTM network for each series to produce next-day predictions. Wavelet meth-
ods decomposed a series into approximation and detail components to better explain be-
haviour over time. He combined these techniques in a novel ensemble model in an at-
tempt to increase forecast accuracy[36].
Zhang, X. proposed a simple but efficient method to predict future stock return rank-
ing without handcrafted features[37].
Achkar, R. proposed an approach to predict stock market ratios using artificial neural
networks. It considered two different techniques BPA-MLP and LSTM-RNN their poten-
tial, and their limitations. And the LSTM-RNN model outperformed the other one
slightly[38].
Zeng, Y. proposed an LSTM model which is based on the dataset SSE50, and the re-
sults showed that the accuracy is above 65%[39].
Shao, X. proposed a Kmeans-LSTM model that used the time window to divide the
stock price sequential data into several equal sub-sequences. And the K-means algorithm
is used to cluster the stock price sub-sequence (He did not give the specific result data.)
[40].
Zhao, Z. proposed a time-weighted LSTM. Unlike other models which treat data in-
discriminately, the proposed model could carefully assign weights to data according to
their temporal nearness towards the data that was used for prediction. And the results
showed that the proposed model outperformed the SVM, RNN, and Random Forest,
model[41].
Nelson, D.M. proposed an LSTM model which was based on the price history and
technical analysis indicators. And the results showed that the proposed model was better
than the MLP and Random Forest model[42].
Dos Santos Pinheiro proposed a character-based neural language model for an event-
based trading model that was combined with NLP and LSTM. The results showed that
the proposed model performed better than some model proposed in other papers such as
WI-RCNN and SI-RCNN[43].
Buczkowski, P. tackled the problem of stock market predicting feasibility, especially
when predictions were based only on a subset of available information, namely: financial
experts' recommendations. The analysis was based on data and results from the ISMIS
2017 Data Mining Competition[44].
Akita, R. proposed an approach that converted newspaper articles into their distrib-
uted representations via Paragraph Vector and modelled the temporal effects of past
events on open prices about multiple companies with LSTM[3].
Chen, K proposed an LSTM model to predict the price of the Chinese stock. But the
results showed that the accuracy is only 27.2%[45].
Qi, L. proposed an LSTM model to predict the price of the forex exchange rate. The
model used technical indicators to calculate events then LSTM is used to make the predic-
tion based on the events. [75]
Pang, Xiong Wen propsed two improved deep LSTM model with embedded layers.
The results showed its improvement over the benchmark. [80]
Feng, Fuli, et al. proposed a noval LSTM model that combined with stock relational
model, it had a great performance compared with its benchmarks. [81]
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 16 of 34

Chung, Hyejung, and Kyung-shik Shin proposed a LSTM model which is optimized
by genetic algorithm (GA). The novel hybrid model had outperformed the benchmark.[82]
Li, Jiahong, et al. proposed a LSTM model that was combined with a Naïve Bayne’s
model. This hybrid model had outperformed benchmarks, and delivered promising re-
sults. [83]
Zhang, Kang, et al. propsed a novel generative adversarial network based on MLP
and LSTM. The model is able to better results than some of the vanilla models. [84]
Jin, Zhigang, et al. propsed to incorporate sentiment analysis model into LSTM, they
successfully created a novel model which deliver a reasonable result. [91]
Long, Jiawei, et al. propsed a hybrid model, which utilised CNN to extract transac-
tional information, then it was passed through a LSTM to predict the stock price. The
model showed improvement over some of the vanilla benchmarks.[92]
Chen,M proposed a LSTM model which uses the sentiment data collected. It com-
bined both sentimental model as well as LSTM to produce a good result. [93]
Qian,Fei. Proposed a LSTM model which used ARIMA to improve its prediction
power. The model yielded a reasonable result.[94]
Li.Z and Tam, V proposed a LSTM model that used wavelet denoising technique
before passing through the LSTM model, the model had produced good result. [95]
4.1.4 DNN
13 articles that used DNN technology for the stock and Forex prediction were briefly
described below, and as shown in the table. Table 5 shows the author, variables, dataset,
and model of all the papers mentioned.
Table 6 The overall information for the papers which used DNN model
# Author Dataset Variables Model
[46] Song, Y. KOSPI and 715 novel input-fea- DNN
KOSDAQ tures(such as (mov-
ing average and dis-
parity of stock price)
[47] Naik, N. NSE SMA, Exponential DNN
ICICI Bank moving average, Mo-
SBI Bank mentum indicator,
Kotak Bank Stochastic oscillator,
Yes Bank Moving average con-
vergence divergence,
Relative strength in-
dex, Williams R, Ac-
cumulation
Distribution index,
Commodity channel
index
[48] Chatzis, S.P. FRED database and Stock price, Ex- DNN
the SNL change rates，VIX
index，Gold price,
TED spread, Oil
price, Effective Fed-
eral Funds Rate,
High yield bond re-
turns
[49] Abe, M. MSCI 25 variables:1.Book- DNN
to-market ratio
2.Earnings-to-price
ratio3.Dividend
yield4.Sales-to-price
ratio5.Cash flow-to-
price ratio 6.Return
on equity 7.Return
on asset
8.Return on invested
capital 9.Accru-
als10.Sales-to-total
assets ratio 11.Cur-
rent ratio12.Equity
ratio13.Total asset
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 17 of 34

growth14.Investment
growth15.Invest-
ment-to-assets ratio
16.EPS Revision(1
month) 17.EPS Revi-
sion(3 months)
18.Market beta19.
Market value 20.
Past stock return(1
month) 21.Past stock
return(12 months)
22 .Volatil-
ity23.Skewness24.Id-
iosyncratic volatil-
ity25. Trading turno-
ver
[50] Nakagawa, K. TOPIX index 16varia- Deep factor
bles:60VOL,BETA,S model(DNN) with
KEW,ROE,ROA,AC layer-wise relevance
CRUALS,LEVER- propagation and mul-
AGE,12- tiple factors)
1MOM,1MOM,60M
OM,PSR,PER,PBR,P
CFR,CAP,ILIQ
[51] Chong, E. KOSPI Close price DNN with autoen-
coder
[52] Singh, R. NASDAQ 36variables:Open (2D²PCA)
price,high price, low +DNN
price, close price,
MA5,MA10,MA20,
BIAS5,BIAS10,DIF
F,BU,BL,K,D,ROC,
TR,MTM6,MTM12,
WR%10,WR%5,OS
C6,OSC12,RSI6,RSI
12,PSY and the deri-
vation of their calcu-
lation
[87] Yu, Pengfei, and S&P 500, DJIA, the Closing price DNN with phase-
Xuesong Yan Nikkei 225 (N 225), space reconstruction
the (PSR) and LSTM
Hang Seng index
(HSI), the China Se-
curities index 300
(CSI 300) and the
ChiNext index
[89] Yong, Bang Xiang, et Singapore stock mar- Intraday prices DNN
al. ket

Song, Y. proposed a DNN model with 715 novel input-features configured on the basis of
technical analysis. He also had proposed a plunge filtering technique to improve the accuracy of
the training model by collecting similar stocks. It was worth noting that the proposed model had
great profitability[46].

Naik, N. proposed a DNN model that used the Boruta feature selection technique to solve the
problem of technical indicator feature selection and identification of the relevant technical indica-
tors. And the results showed that his model performed much better than ANN and SVM model[47].

Chatzis, S.P. proposed a DNN model in which it used Boosted approaches to predict stock
market crisis episodes. According to his research, it was meaningful to know the stock market crisis
to predict the price, even though his research was not specific to certain prediction methods[48].
Abe, M. proposed a DNN network and his results showed that DNNs generally outperform
shallow neural networks, and the best networks also outperformed representative machine learning
models[49].
Nakagawa, K. proposed a deep factor model and a shallow model with DNN. The deep factor
model outperformed the linear model. This implied that the relationship between the stock returns
in the financial market and the factors is nonlinear, rather than linear. The deep factor model also
outperformed other machine learning methods including SVR and random forest. The shallow
model was superior in accuracy, while the deep model was more profitable[50].
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 18 of 34

Chong, E. proposed DNN networks and examined the effects of three unsupervised feature
extraction methods including principal component analysis, auto-encoder, and the restricted Boltz-
mann machine, on the network's overall ability to predict future market behaviour. Empirical results
suggested that DNNs could extract additional information from the residuals of the auto-regressive
model and improve prediction performance; the same could not be said when the auto-regressive
model is applied to the residuals of the network[51].
Singh, R. proposed a 2-Directional 2-Dimensional Principal Component Analysis
(2D²PCA)+DNN which outperformed than the RNN and (2D²PCA) +RBFNN. The paper found
that the best results were generated from a window size of 20 and a dimension of 10 × 10. The
deep learning method for higher dimensions and large window sizes was giving a limited perfor-
mance[52].
Yu, Pengfei, and Xuesong Yan proposed a novel DNN model which incorporated LSTM as
well as phase-space recognition. The model had produce promising results. [87]
Yong, Bang Xiang, et al. propsed a DNN model with 40 nodes which showed a rea-
sonable results and appear to be a highly profitable model. [89]

4.1.5 Reinforcement learning

8 articles that used reinforcement learning technology for the stock and Forex predic-
tion were briefly described below, and as shown in the table. Table 6 showed the author,
variables, dataset, and model of all the papers mentioned.
Table 6 The overall information for the papers which used Reinforcement learning model
# Author Dataset Variables Model
[57] Li, Y. US stock dataset Close price and Vol- DQN
ume
[58] Shin, H.-G. KOSPI a candlestick chart, Reinforcement
four moving average Learning combined
curves (5, 20, 60, with LSTM and
and 120 days), CNN
, a bar graph of trad-
ing volume, DMI,
and SSO
[59] Jia, W. Chinese stock codes: Open price,high Reinforcement
002415, 600016, price, low price, Learning with
600028, 600547, close price, vol- LSTM-based agent
600999 and 601988 ume,DEA ,MACD
EXPMA,CDP ,TRIX
,BBI,
ASI ,KDJ ,RSI ,
PSY
VR ,ADX ,CCI ,WR
,dm up,dm down
[60] Carapuço, J. EUR/USD bid/ask prices and Reinforcement
volumes Learning
[61] Kang, Q. S&P500 index open, low, high, Reinforcement
close price and trad- Learning with A3C
ing volume algorithm
[62] Zhu, Y. S&P500 index open, low, high, Reinforcement
close price ,vol- Learning with ABN
ume,MACD,MA6,M
A12,RSI6,RSI12 and
KD
[63] Si, W. Stock-IF,stock- Close price multi-objective deep
IC,stock-IF reinforcement learn-
ing with LSTM
[64] Pastore, A. FTSE100 stock in- Date, Type, Stock, Reinforcement learn-
dex Volume, Price, and ing
Total

Li, Y. proposed three different reinforcement learning methods to predict the price of the
stock. The results showed that the best-performing Deep Reinforcement Learning model is DQN,
not Double DQN. The paper also demonstrated Duelling DQN, which was an improved model
based on DQN[57].
Shin, H.-G. proposed a Reinforcement Learning model that combined with LSTM and CNN.
The model generated various charts from stock trading data and used them as inputs to the CNN
layer. The features extracted through the CNN layer were divided into column vectors and inputted
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 19 of 34

to the LSTM layer. The reinforcement learning defines agents' policy neural network structure,
reward, and action, and provides buying, selling, and holding probabilities as final output[58].
Jia, W. proposed a Reinforcement Learning with an LSTM-based agent who could automati-
cally sense the dynamics of the stock market and alleviate the difficulty of manually designing
indicators from massive data. The paper had compared a wide range of different input sets[59].
Carapuço, J. proposed a reinforcement learning-Q network model, Three hidden layers of
ReLU neutrons were trained as RL agents through the Q-learning algorithm under a novel simu-
lated market environment framework. The framework was able to consistently induce stable learn-
ing that generalised to out-of-sample data[60].
Kang, Q. proposed to apply the state-of-art Asynchronous Advantage Actor-Critic algo-
rithm(A3C algorithm) to solve the portfolio management problem and designed a standalone deep
reinforcement learning model[61].
Zhu, Y. proposed an adaptive box-normalisation (ABN) stock trading strategy based on rein-
forcement learning, which improved the original box theory. In his ABN strategy, the stock market
data was independently normalised inside each oscillation box[62].
Si, W. proposed a reinforcement learning model which had multi-objective and LSTM agent.
It could be found that feature learning could contribute to better performances. The LSTM network
made continuous decisions and could change positions at the right time, which reduced the trans-
action cost, and the multi-objective structure made good profits within the acceptable risk[63].
Pastore, A. analysed data for 46 players extracted from a financial market online game and
tested whether Reinforcement Learning could capture these players' behaviour using a riskiness
measure based on financial modelling[64].

4.1.6 Other Deep Learning Methods

8 articles that used other deep learning technology for the stock and Forex prediction
were briefly described below, and as shown in the table. Table 7 showed the author, var-
iables, dataset, and model of all the papers mentioned.
Table 7 The overall information for the papers which used other deep learning methods
# Author Dataset Variables Model
[65] Long, W. CSI300 Open price, high MNFF
price, low price,
close price, Volume
[66] Wu, J.-L. ANUE stock message as in- HAN
formation to form
the text feature of
each stock news(ti-
tle, summary, and
keywords)
[67] Cho, C.-H. CATHAY HOLD- Open price,high Wavenet
INGS, Fubon Finan- price, low price,
cial, CTBC HOLD- close price, vol-
INGS, ESFH, and ume,MACD,CCI,AT
FFHC R,BOLL,EMA12/20,
MA5,MA1MOM6,
MOM12,ROC,RSI,
WVAD,Exchange
rate and Interest rate
[68] Minh, D.L S&P 500, VN-index Open price, high Documents prepro-
and cophieu68; price, low price, cessing-Documents
Bloomberg, Reuters close price, Volume, labeling -Stock2Vec
Stochastic oscillator, embedding-BGRU
William (%R), and
RSI
Processed news arti-
cle
[69] Hu, G. Financial Times Candlestick Convolutional Auto-
Stock; Exchange 100 charts(images rather Encoder(CAE)
Index (FTSE 100) than annotation data)
[70] Hu, Z. Chinese stock price; 1.Close price and HAN with SPL
News(Not given spe- volume; 2.news cor-
cific data) pus sequence
[72] Kim, T. and Khushi Nine Dow Jones Open price, high 2D Relative-Atten-
companies represent- price, low price, tional Gated Trans-
ing each sector - in- close price, Volume former
dustrials (MMM), fi-
nancials (JPM),
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 20 of 34

consumer services
(PG), technology
(AAPL), health care
(UNH), consumer
goods (WMT), oil &
gas (XOM), basic
materials (DD) and
telecommunications
(VZ) From Yahoo.
[74] Zhang, and Khushi Forex Exchange Trend indicators: Genetic Algorithm
rates Moving average, Ex-
ponential moving av-
erage, Double expo-
nential moving aver-
age, Triple exponen-
tial moving average,
Vortex indicators
Momentum indica-
tors: Relative
strength index, Sto-
chastic oscillators,
Volatility indicators:
Bollinger bands,
Ichimoku indicators
[79] Hu, Ziniu Chinese stiock data News and financial Hybird attention net-
and economic news data works
[86] Shi, Lei, et al. Apple Inc. and S&P News and financial Hybrid of RNN,
500 data LSTM and CNN

Long, W. proposed a novel end-to-end model named multi-filters neural network

(MFNN) specifically for feature extraction on financial time series samples and price
movement prediction task. Both convolutional and recurrent neurons are integrated to
build the multi-filters structure. And the results showed that the proposed model is better
than the RNN, CNN, and traditional machine learning methods[65].
Wu, J.-L. proposed a keyword-based attention network into Hierarchical Attention
Networks (HAN), namely the HKAN model, to learn the relationships between dimen-
sional sentiments (trend and trading) and stock messages which outperformed the HAN
network[66].
Cho, C.-H.proposed three different models to predict the price of the stock, they
were: LSTM，Seq2seq and Wavenet. According to his experiments, Wavenet outper-
formed the other two models[67].
Minh, D.L proposed a network to predict the directions of stock prices by using both
financial news and sentiment dictionary. His results showed that TGRU achieved better
performance than GRU and LSTM and Stock2Vec is more efficient in dealing with finan-
cial datasets[68].
Hu, G. proposed a Convolutional AutoEncoder model to learn a stock representation
and converted a 4-channel stock time series (lowest, highest, open, and close price for the
day) to candlestick charts by using synthesis technique to present price history as images.
This method had successfully avoided expensive annotation. And the proposed model
outperformed the FTSE 100 index and many well-known funds in terms of total re-
turn[69].
Hu, G. proposed a Hybrid Attention Networks (HAN) to predict the stock trend
based on the sequence of recent related news. Moreover, he applied the self-paced learn-
ing(SPL) mechanism to achieve effective and efficient learning. The results showed the
proposed model outperforms than RNN, Temporal- Attention-RNN, News-Attention-
RNN, and HAN[70].
Kim, T. proposed a 2D relative-attentional gated transformer which was used to op-
timise portfolio return. It used general reinforcement learning with the agent incorporat-
ing 2D Relative-attentional Gated Transformer. [72]
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 21 of 34

Zhang, Z. proposed a genetic algorithm using a crossover of technique indicators as

input. It had successfully outperformed some of the traditional trading strategies.[74]
Hu, Ziniu proposed hybrid attention networks which is a novel way of combining
financial time-series information and news nature language information. [79]
Shi, Lei, et al. propsed a system where it perform factor analysis then utilise multiple
depp learning method to design a model which outperformed benmark. [86]

4.2 Papers Results grouped by the method used

4.2.1 CNN
Since the performance metrics used in different articles were different, our survey
analysed them based on various metrics. Results used were the average performance of
the best performing models in the mentioned papers. Table 8 showed the metrics and
results for the papers which used the CNN model.
Table 8 The metrics and results for the papers which used the CNN model
Performance Metrics Reference no. Corresponding value Performance Metrics Reference no. Corresponding value

RMSE [8] 0.043+/- .007 MAPE [9] 5

[9] 11 Sharpe ratio [11] [11]2D:0.1422, 3D:0.1413
[71] 0.395185*10^-3
MAE [9] 6
[15] [15]0.611
[71] 0.240977*10^-3
Accuracy [10] 71% CEQ [11] 2D:0.0006681 3D:0.000664
[13] 60.02% Return rate [11] [11](2D:1.2312
[16] 55.44%
[19] 71.72% 3D:1.2604)
[20] 65.08% [13] [13](1.3107)
[76] 75.2%
[77] 78.46%
[78] 57.88%
[90] 74.753%
Error Percentage [17] 95.02% [15] [15](1.2156)
F-measure [11] 2D:0.4944 [18] [18](1.309)
3D:0.4931 Mean Test Score [12] 0.000281317
[15] 0.6227 MSE [19] 0.2631
[71] 0.156*10^-6
[16] 0.7133 AE [8] 0.029+/-.005
[76] 0.73
[90] 0.6367

4.2.2 RNN
Since the performance metrics used in different articles were different, our survey
analysed them based on different metrics. Results used were the average performance of
the best performing models in the mentioned papers. Table 9 shows the metrics and re-
sults for the papers which used the RNN model.
Table 9 The metrics and results for the papers which used RNN model
Performance Metrics Reference no. Corresponding value
RMSE [53] 512-530
[55] 0.0205
[73] 0.00165
MAPE [55] 0.2431
[73] 0.232
MAE [55] 0.0132
Accuracy [54] 68.95%
[55] 66.54%
F-measure [54] 0.7658
Recall [54] 0.7471
Precision [54] 0.7855
MSE [56] 0.057443

4.2.3 LSTM
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 22 of 34

Since the performance metrics used were different in different articles, our survey
would analyse them based on different metrics. Results used were the average perfor-
mance of the best performing models in the mentioned papers. Table 10 showed the met-
rics and results for the papers which used the LSTM model.
Table 10 The metrics and results for the papers which used LSTM model
Performance Met- Reference no. Corresponding Performance Met- Reference no. Corresponding
rics value rics value
RMSE [21] 0.306543 MAPE [24] 1.03
[24] 347.46 [27] 1.6905
[25] 0.0151 [29] 4.13
[26] 25.90 [30] 1.0077
[75] 0.119
[82] 0.91
[84] 1.37
[91] 1.65
[95] 0.6346
[27] 0.0242 Precision [42] 0.553
[28] 1.3 Recall [42] 0.129
[34] 9.72 Return rate [22] 1.0667
[35] 4.24(Average) MSE [21] 0.093969
[36] 1-10 [22] 0.004845492
[75] 0.0015
[94] 0.02295
[81] 0.000379
MAE [21] 0.21035 [24] 120731.4
[24] 262.42 [27] 19.7096
[26] 0.1895 [28] 0.019
[27] 0.0169 [29] 0.00098
[29] 0.023 [30] 7.56
[31] 0.01069 [31] 0.00149
[30] 1.975 [32] 1.012
Accuracy [23] 54.58% MCC [23] 0.0478
[25] 98.49% R2 [24] 0.83
[34] 60.60% HMAE [31] 0.42911
[39] 65% HMSE [31] 0.23492
[41] 83.91% IC [37] 0.1259
[42] 55.90% AR [37] 0.2015
[43] 63.34% IR [37] 3.0521
[45] 27.20%
[80] 53.2%
[83] 87.86%
[91] 70.56% Score [44] 0.4271
[92] 75.89%
[93] 75.58%
F-measure [42] 0.209

4.2.4 DNN
Since the performance metrics used in different articles were different, our survey
analysed them based on different metrics. Results used were the average performance of
the best performing models in the mentioned papers. Table 11 showed the metrics and
results for the papers which used the DNN model.
Table 11 The metrics and results for the papers which used DNN model
Performance Met- Reference no. Corresponding Performance Met- Reference no. Corresponding
rics value rics value
RMSE [50] 0.0951 Sharpe ratio [50] 1.41
[89] 5.34
[51] 0.8214 Return rate [49] 1.0952
[52] 0.00674 [50] 1.1081
MAE [50] 0.0663 CORR [49] 0.0582
[51] 0.5852 MSE [49] 0.0836
Accuracy [46] 61.90% [51] 0.9621
[47] 84.50% SMAPE [52] 0.0696
[87] 58.07%
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 23 of 34

MAPE [52] 0.080059

F-measure [47] 0.824 [89] 1.84
Volatility [50] 7.65%

4.2.5 Reinforcement learning

Since the performance metrics used were different in different articles, our survey
would analyse them based on different metrics. Results used were the average perfor-
mance of the best performing models in the mentioned papers. Table 12 showed the met-
rics and results for the papers which used the Reinforcement learning model.
Table 12 The metrics and results for the papers which used Reinforcement learning model
Performance Metrics Reference no. Corresponding value
Sharpe ratio [58] 2.77
[63] 0.12
Return rate [59] 1.948
[60] 1.163 ± 2.8%
[62] 2.442
MSE [62] 0.000412

4.2.6 Other Deep Learning Methods

Since the performance metrics used in different articles were different, our survey analysed
them based on different metrics that were used. Results used were the average performance
of the best performing models in the mentioned papers. Table 13 showed the metrics and
results for the papers which used other deep learning methods.
Table 13 The metrics and results for the papers which used other deep learning methods
Performance Metrics Reference no. Corresponding value
RMSE [67] 0.6866
Accuracy [68] 66.32%
[70] 47.8%
[86] 79.7%
Sharpe ratio [65] 4.49.
[69] 0.8
[72] 0.6418
[74] 6.68 on EURUSD currency
Return rate [65] 1.4228
[68] 1.0531(0.25%)
[69] 1.118
[70] 1.611(0.3%)
[72] 1.4316
[74] 1.0968 on EURUSD currency
[79] 1.52
MSE [66] 1.05
MDAE [66] 0.71
Correlation [67] 0.9564
Precision [68] 72.1%
Recall [68] 77.32%

5. Discussion
5.1 Analysis based on the method used
5.1.1 CNN
Some findings can be drawn from reviewing the CNN models:
1. According to the datasets that all the paper used, 6 papers used the combination
of technical analysis and sentiment and news analysis to predict the stock. The rest of
them used the method of technical analysis only.
2. As for the variables, the close price was the choice of all CNN models, and five
papers were using close price only.
3. It could be found that 12 of the papers changed the traditional CNN model to pur-
sue higher performance of prediction. And the combination of CNN and LSTM was the
most common model.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 24 of 34

4. The metrics used in each paper were different; there were 11 metrics used for meas-
uring the performance of the CNN model.
5.Multiple articles selected RMSE, Return rate, F-measure, Sharpe ratio, and accu-
racy. We could find that the paper [19] had the highest accuracy than any other papers,
paper [13] had the highest return rate followed by the paper [18], [11] and [15]. Paper [71]
had the lowest RMSE. Paper [16] had a higher F-measure than paper [11], and [15], but
paper [15] achieved a higher Sharpe ratio than paper [11].

5.1.2 RNN
In the section of the RNN based model, the following conclusions could be drawn:
1. According to the datasets that all the paper used, there was 1 paper that used the
combination of technical analysis and sentiment and news analysis to predict the stock.
And the rest of them used technical analysis only.
2. For the variables, all the RNN based models used the multivariable input and open
price, close price, highest price, and the lowest price are used in all models as an input.
3. It could be found that all the papers changed the traditional RNN model to pursue
higher performance of prediction. Two of the papers chose the C-RNN based model.
4. The metrics used in each paper were different; in total, there were 8 metrics used
for measuring the performance of the RNN based model.
5.RMSE and accuracy were selected by multiple articles. We could note that paper
[55] has a much lower RMSE than paper [53], and paper [54] had higher accuracy than
paper [55].

5.1.3 LSTM
Following points were worth discussing from reviewing the LSTM papers:
1. According to the datasets that all the paper used, 3 papers used a combination of
technical analysis and sentiment and news analysis to predict the stock. The rest of them
used technical analysis only with the exception of one paper which used expert recom-
mendations.
2. As for the variables, the close price was the choice of 23 LSTM based models, there
were 8 papers using close price only, and 12 papers selected to include close price, open
price, high price, and low price in their input.
3. It could be found that 15 of the papers changed the traditional LSTM model to
pursue higher performance of prediction. Attention-based LSTM and LSTM with RNN
were the most frequent models that showed up in three different papers. Two papers
chose the method of LSTM with GRU to improve the model.
4.The metrics used in each paper were different; there were 17 metrics used for meas-
uring the performance of the LSTM based model.
Multiple articles selected 5.RMSE, MAPE, MAE, accuracy, and MSE, we could find
that Paper [25], paper [94] and paper [27] was in the lowest order of magnitude with paper
[25] achieving the lowest RMSE; paper [21] was in the second-lowest order of magnitude;
paper [28],[35],[36] and paper [34] were in the third lowest order of magnitude. Paper [26]
was in the fourth lowest order of magnitude and paper [24] had the highest order of mag-
nitude.
6. As for MAPE, paper [75] had the lowest order of magnitude followed by the paper
[30] ,[24], [82], [84],[91],[95] and [27]. And paper [29] had the highest order of magnitude
of MAPE.
7. As for MAE, paper [75], [31], [27] and [29] had the lowest order of magnitude MAE
with paper [75] having the best performance; paper[26] and [21] were in the second-lowest
order of magnitude while paper [30] was in the third lowest order of magnitude. Paper
[24] had the highest order of magnitude.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 25 of 34

8. As for MSE, paper [81]had the lowest MSE and was in the lowest order of mag-
nitude; paper [29] , [31] and [22] were in the second-lowest order of magnitude while pa-
per [28] and [21] were in the third lowest order of magnitude. Paper [32] and [30] were in
the fourth lowest order of magnitude, paper [27] was in the fifth-lowest order of magni-
tude, and paper [24] had the highest order of magnitude of MAE.
9.As for accuracy, paper [25] had the highest accuracy followed by [83], [41],[39], [43],
[34], [42], [23] and paper [45] had the lowest accuracy.

5.1.4 DNN
In the section of DNN-based model, the following conclusions could be drawn:
1. According to the datasets that all the paper used, no paper used the combination
of technical analysis and sentiment and news analysis to predict the stock. All of them
used the method of technical analysis only.
2. As for the variables, six of the seven DNN-based models used multivariable input,
and only one paper used close price as its sole input.
3. It could be found that 3 of the papers changed the traditional DNN model to pursue
higher performance of prediction. All of the improved models were not duplicated.
4. Because the metrics used in each paper were different; there were 11 different met-
rics used for measuring the performance of the DNN-based model.
5. Multiple articles selected 5.RMSE, MAE, accuracy, return rate, and MSE. We could
find that the paper [52] had the lowest RMSE followed by the paper [50] and [51], with
the latter two in different orders of magnitude.
6. As for MAE, paper [50] had a lower MAE than paper [51], and the accuracy of the
paper [47] was higher than paper [46] and [87]. Furthermore, paper [50] had a higher re-
turn rate than paper [49]. And paper [49] had a lower MSE than paper [51].

5.1.5 Reinforcement learning

In the section of Reinforcement learning-based model, the following conclusions
could be drawn:
1. According to the datasets that all the paper used, no paper used the combination
of technical analysis and sentiment and news analysis to predict the stock. All of them
used technical analysis only.
2. As for the variables, 7 of the reinforcement learning-based models used the multi-
variable input, and only one paper solely used close price as input.
3. It could be found that three of 6 papers changed the traditional Reinforcement
learning model to pursue higher performance of prediction. 3 of the improved models
were combined with LSTM.
4. The metrics used in each paper were different; in total, there were 3 metrics used
in the measurement performance of the Reinforcement learning-based model.
5. Multiple articles selected 5. Sharpe ratio and return rate, we could find that paper
[58] had a much higher Sharpe ratio than paper [63]; paper [62] had a higher return rate
than paper [59] then followed by the paper[60].

5.1.6 Other Deep Learning Methods

In the section of other deep learning methods based model, the following conclusions
could be drawn:
1. According to the datasets that all the paper used, 4 papers made use of sentiment
and news analysis to predict the stock. The rest of them used the method of technical
analysis only.
2. As for the variables, five of the other deep learning methods models used the mul-
tivariable input and only one paper used candlestick charts alone as input.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 26 of 34

3. It could be found that there were 5 different models in the other deep learning
methods section. The only model that appeared three times was HAN, which consist of
an ordinary HAN model and two modified HAN models. The rest of the models were not
duplicated.
4. Because the metrics used in each paper were different, there were 9 metrics used
in the measurement performance of this section.
5. Multiple articles selected 5. Accuracy, Sharpe ratio, and Return rate, we could find
that the paper [86] had the highest accuracy.Paper [65] had a much higher Sharpe ratio
than paper [69].
6. As for the return rate, paper [70] had the highest return rate followed by the paper
[65], [69]and paper [68] had the lowest return rate.

5.2 Discussion and Analysis based on performance metrics

In this part, all of Forex/Stock price prediction models mentioned above, which made
use of the deep learning, would be discussed and analysed. All analysis of the perfor-
mance metric data was shown in Table 14. It could be found that the most commonly used
performance metrics were RMSE, MAPE, MAE, MSE, Accuracy, Sharpe ratio, and Return
rate. Hence our review would analyse the papers in terms of its performance metrics used.
Table 14 Analysis based on Performance Metrics
Performance Metrics Reference no.
RMSE [8], [9], [21], [24], [25], [26], [27], [28], [34],
[35],[36],[50], [51], [52], [53], [55], [67],[71],[73]
MAPE [9], [24], [27], [29], [30], [52], [55],[73],[82],[84],[94]
MAE [9], [21], [24], [26], [27], [29], [31], [30], [50], [51],
[55],[71],[95]
Accuracy [10],[13],[16], [19], [20],[23], [25], [34],[39], [41], [42],
[43], [45],[46], [47], [54], [55], [68],
[70],[76],[77],[78],[80],[83],[87],[90],[91],[92],[93]
F-measure [11], [42], [47], [54],[15],[16],[76],[90]
Sharpe ratio [11], [15],[50],[58], [63], [65], [69], [72],[74]
CEQ [11]
Return rate [11], [13],[15],[18], [22], [49], [50],[59], [60],
[65],[68],[69],[70] , [72],[74],[79]
Mean Test Score [12]
MSE [19], [21], [22], [24], [27], [28], [29], [30], [31], [32], [49],
[51],[56],[62],[71],[81]
AE [8]
Precision [42], [54]
Recall [42], [54]
R2 [24]
Error Percentage [17]

MCC [23]
HMAE [31]
HMSE [31]
CORR [49]
SMAPE [52]
Volatility [50]
IC [37]
AR [37]
IR [37]
Score [44]

5.2.1 analysis based on RMSE

In this part, the value of the RMSE in the papers would be discussed. Table 15 showed
the details of each paper’s RMSE based on the numerical range. In this table, paper [8],
[9], [71] used CNN based model; paper [21], [24], [25], [26], [27], [28], [34],[35],[36],
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 27 of 34

[75],[94] used LSTM based model; paper [50], [51], [52] used DNN based model; paper
[53], [55], [73] used RNN based model and paper [67] used other deep learning method.
Table 15 Analysis based on RMSE
RMSE Reference no.
range
RMSE [8], [9], [21], [24], [25], [26], [27],
[28], [34], [35],[36],[50], [51],
[52], [53], [55], [67],[71],[75],[94]
<0.001 [71], [75]
0.001-0.01 [52], [73]
0.01-0.1 [8], [21], [25],[27],[50], [55],[94]
0.1-1 [9],[51],[67]
1-10 [28],[34],[35],[36]
10-100 [26]
>100 [24], [53]

It was clear that paper [71], [75] achieved the best performance using DNN model.
Papers that had a RMSE smaller than 0.001, paper [52], [73], [8], [21], [25], [27], [50], [55]
had great performance while paper [26], [24], [53] didn’t have the low RMSE .
Among all the papers, 33% of the CNN papers had an RMSE below 0.001, 33% of the
CNN papers were in the range of 0.01-0.1, and 33% of the CNN papers were in the range
of 0.1-1. 36% of the LSTM papers were in the range of 0.01-0.1 and 1-10 respectively, , the
rest LSTM papers were distributed equally in range of <0.001, 10-100 and above 100 with
9%in each range. 33.3% of the DNN papers were in the range of 0.001-0.01, 0.01-0.1 and
0.1-1 respectively. The RNN papers were distributed equally in the range of 0.001-0.01,
0.01-0.1 and above 100, i.e. 33% in each range. The only paper that used other deep learn-
ing method had an RMSE in the range of 0.01-0.1.

5.2.2 analysis based on MAPE

In this part, the value of the MAPE would be discussed. Table 16 showed the papers'
MAPE performance based on the numerical range. In this table, paper [9] used CNN based
model; paper [24], [27], [29], [30], [75], [82], [84], [95] used LSTM based model; paper
[52],[89] used DNN based model; paper [55] used RNN based model.
Table 16 Analysis based on MAPE
MAPE range Reference no.
MAPE [9], [24], [27], [29], [30], [52],
[55],[75],[82],[84],[89], [95]
0-0.5 [52],[55],[75], [95]
0.5-1 [82]
1-1.5 [24],[30],[84],[89]
1.5-2 [27]
2-10 [9],[29]

It could be found that paper [52], [55],[75] performed best which used the DNN
model and RNN model and LSTM respectively, and both were in the range of 1-1.5, paper
[24],[30],[82] had great performance while paper [9], [29] didn't have a low MAPE.
Among all the papers, the only CNN paper was in the range of 2-10. 37.5% of the
LSTM papers were in the range of 1-1.5, 12.5% of the LSTM papers were in the range of
0.5-1, 1.5-2, and 2-10 respectively, 25% of LSTM papers in a range of 0-0.5. The two DNN
papers were in the range 0-0.5, and 1-1.5. Lastly, the only one RNN paper was in the range
of 0-0.5.

5.2.3 analysis based on MAE

In this part, the value of the MAE would be discussed. Table 17 showed the details
of papers’ MAE. In this table, paper [9], [71] used CNN based model; paper [23], [26],
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 28 of 34

[28], [29], [31], [33], [32] used LSTM based model; paper [50], [51] used DNN based model
and paper [55] , [73] used RNN based model.
Table 17 Analysis based on MAE
MAE range Reference no.
MAE [9], [21], [24], [26], [27], [29],
[31], [30], [50], [51], [55],
[71]
<0.01 [71]
0.01-0.1 [27],[29],[31],[50],[55]
0.1-1 [21],[26],[51],[73]
1-10 [9],[30]
10-100 N/A
>100 [24],

It could be found that paper [71] performed best which used CNN model. Papers'
MAE were in the range of 0.1-1. Paper [27], [29], [31], [50], [55] also had good performance,
however, paper [24] didn’t have a low MAE.
Among all the papers that used MAE as measurement, 50% of CNN paper was in the
range of 1-10; the other 50% had an MAE small than 0.01. 42.9% of the LSTM papers were
in the range of 0.01-0.1, 28.6% of the LSTM papers were in the range of 0.1-1, 14.2% of the
LSTM papers were in the range of 1-10 and above 100 respectively, 50% of the DNN pa-
pers were in the range of 0.01-0.1 and 0.1-1 respectively. The two RNN papers were in the
range of 0.01-0.1 and 0.1-1.

5.2.4 analysis based on MSE

In this part, the value of the MSE would be discussed. Table 18 showed the details of
papers’ MSE. In this table, paper [19], [71] used CNN based model; paper [21], [22], [24],
[27], [28], [29], [30], [31], [32], [81] used LSTM based model and paper [49], [51] used DNN
based model. Paper,[56] used RNN based model, and paper [62] used the Reinforcement
learning based model.
Table 18 Analysis based on MSE
MSE range Reference no.
MSE [19], [21], [22], [24], [27], [28],
[29], [30], [31], [32], [49],
[51],[56],[62],[71]
<0.01 [71],[81]
0-0.01 [22],[29],[31],[62]
0.01-0.1 [21],[28],[49],[56]
0.1-1 [19],[32],[51]
1-10 [30]
10-100 [27]
>100 [24]

It could be found that paper [81] performed the best which used the LSTM model; its
MSE was smaller than 0.01. Paper [71], [22], [29], [31],[62] also had good performance,
however, paper [24], [27] didn’t have the low MSE.
Among all the papers that used MSE as performance measurement, the only CNN
paper was in the range of 0.1-1, 40% of the LSTM papers were in the range of 0-0.01, 20%of
the LSTM papers were in the range of 0.01-0.1, 10% of the LSTM papers were in the range
of 0.1-1, 1-10, 10-100 and above 100 respectively, 50% of the DNN papers were in the range
of 0.01-0.1 and 0.1-1 respectively. The only RNN based model and reinforcement learning-
based model was in the range of 0-0.01 and 0.01-0.1 respectively.

5.2.4 analysis based on the accuracy

In this part, the value of the accuracy would be discussed. Table 19 shows the details
of papers’ accuracy. In this table, paper [10], [13],[16], [19],[20],[76],[77],[78],[90] used
CNN based model; paper [23], [25], [34], [39],[41], [42], [43],[45], [80], [83], [91],[92],[93]
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 29 of 34

used LSTM based model;paper [46], [47], [87] used DNN based model; paper [54], [55]
used RNN based model and paper [68], [70],[86] used other deep learning method.
Table 19 Analysis based on the accuracy
Accuracy range Reference no.
Accuracy [10],[13],[16], [19], [20],[23], [25],
[34],[39], [41], [42], [43], [45],[46], [47],
[54], [55], [68],
[70],[76],[77],[78],[80],[83],[86],[87],
[90],[91],[92],[93]
0-50% [45],[70]
50%-60% [16],[23],[42],[78],[80],[87]
60%-70% [13], [20],[34],[39],[43],[46],[54],
[55],[68]
70%-80% [10],[19],[76],[77],[86],[90],[91],[92],[93]
80%-90% [41],[47].[83]
90%-100% [25]

It could be found that paper [25] performed best, which used LSTM model, and the
paper's accuracy was in the range of 90%-100%. Paper [41], [47] ,[83] also had great per-
formance, but paper [45], [70] didn’t have the high accuracy.
Among all the papers that had accuracy measure metric, 22% of the CNN papers
were in the range of 50%-60%, and 22% of the CNN papers were in the range of 60%-70%
and 55% of the CNN papers were in the range of 70%-80%. 8% of the LSTM papers were
in the range of 0-50%, and 90%-100% respectively,15% of the LSTM papers were in the
range of 80%-90%, 23% of the LSTM papers were in the range of 50%-60%, 60%-70% and
70%-80% respectively. 33% of the DNN papers were in the range of 50%-60%, 60%-70%
and 80%-90% respectively. All the RNN papers were in the range of 60%-70%. 33% of
other deep learning methods were in the range of 40%-50%, 60%-70% and 70-80% respec-
tively.

5.2.5 analysis based on Sharpe ratio

In this part, the value of the Sharpe ratio would be discussed. Sharpe ratio referred
to the amount of return generated relative to the risk taken, where risk was calculated
using the standard deviation of return. Table 20 showed the details of the papers' that
used the Sharpe ratio as a preformance metric. In this table, paper [11] and [15] used
CNN based model; paper [50], [89] used DNN based model; paper [58], [63] used Rein-
forcement learning-based model, and paper [65], [69], [72], [74] used other deep learning
method.
Table 20 Analysis based on Sharpe ratio
Sharpe ratio range Reference no.
Sharpe ratio [11], [15],[50],[58], [63], [65],
[69],[72],[74]
0.1-1 [11], [15],[63],[69],[72]
1-2 [50]
2-5 [58]
5-10 [65],[74],[89]

It could be found that paper [65], [74],[89] performed best, which used DNN model,
and Sharpe ratio was in the range of 5-10. Paper [58] also had a great performance, but
paper [11], [15],[63], [69], [72] didn’t have the high Sharpe ratio.
Among all the paper that used the Sharpe ratio, all CNN papers were in the range of
0.1-1. The two DNN papers were in the range of 1-2 and 5-10. 50% of the Reinforcement
learning paper was in the range of 0.1-1, and 50% of the Reinforcement learning paper
was in the range of 2-5. 50% of the other deep learning method paper was in the range of
0.1-1, and 50% of the other deep learning method paper was in the range of 5-10.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 30 of 34

5.2.6 analysis based on the Return rate

In this part, the value of the return rate would be discussed. Table 21 showed the
details of papers’ return rate. In this table, paper [11],[13],[15], [18] used CNN based
model; paper [22] used LSTM based model; paper [49], [50] used DNN based model; pa-
per [59], [60] used Reinforcement learning based model and paper [65], [68], [69], [70],
[72], [74], [79] used other deep learning method.
Table 21 Analysis based on the return rate
Return rate range Reference no.
Return rate [11], [13],[15],[18], [22], [49],
[50],[59], [60], [65],[68],[69],[70]
1.0-1.2 [22], [49], [50],[60],[68],[69], [74]
1.2-1.4 [11],[13],[15],[18]
1.4-1.6 [65],[72],[79]
1.6-1.8 [70]
1.8-2.0 [59]

It could be found that paper [59] performed the best which used Reinforcement learn-
ing model, and the paper's return rate was in the range of 1.4-1.8. Paper [65] and [70] had
good performance but paper [22], [49], [50], [60], [68], [69], [74] didn’t have the high return
rate.
Among all the papers that used return rate as a performance measurement, all the
CNN papers were in the range of 1.2-1.4, and the only LSTM paper was in the range of
1.0-1.2, all the DNN papers were in the range of 1.0-1.2. 50% of the Reinforcement learning
papers were in the range of 1.0-1.2, and 50% of the Reinforcement learning papers were
in the range of 1.8-2. The other deep learning method papers performed well with 14%
paper in the range of 1.0-1.2, 42% paper in the range of 1.4-1.6 , and 42% paper in the range
of 1.6-1.8.

6. Conclusions
This paper provided a detailed review of 88 papers from 2015 to the present on pre-
dicting stock /Forex price movements through deep learning methods. The existing
stock/Forex models were evaluated through analysing data sets, variables, the use of dif-
ferent models, and the metrics of evaluation. The research review included a wide range
of techniques: CNN, LSTM, DNN, RNN, Reinforcement learning, and other deep learning
methods such as HAN, NLP, and Wavenet. Furthermore, the data sets, variables, models,
and their different results were analysed and compared within each technique. Then our
paper discusses the main performance metrics of all models. They are RMSE, MAPE,
MAE, MSE, Accuracy, Sharpe ratio and Return rate.
This paper aimed to contribute to the research of stock/ Forex market prediction
through the analysis of the above different deep learning prediction models. Through the
review, It can be identified that there is a lack of studies on the combination of multiple
deep learning methods, espically in respect to other deep learning methods. The hybrid
networks are showing promising signs for future research. In the future, we would design
a specific hybrid model based on the above analysis, incorporating latest technology such
as advanced genetic algorithms, self attention neural networks to predict the stock/Forex
market.
Author Contributions: First two authors contributed equally, MK conceived & designed the study
supervised the work and revised the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.

References
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 31 of 34

1. Tsai, C.-F. and Y.-C. Hsiao, Combining multiple feature selection methods for stock prediction: Union, intersection, and
multi-intersection approaches. Decision Support Systems, 2010. 50(1): p. 258-269.
2. Nassirtoussi, A.K., et al., Text mining of news-headlines for FOREX market prediction: A Multi-layer Dimension Reduc-
tion Algorithm with semantics and sentiment. Expert Systems with Applications, 2015. 42(1): p. 306-324.
3. Akita, R., et al. Deep learning for stock prediction using numerical and textual information. in 2016 IEEE/ACIS 15th Inter-
national Conference on Computer and Information Science (ICIS). 2016. IEEE.
4. Sirignano, J. and R. Cont, Universal features of price formation in financial markets: perspectives from deep learning.
Quantitative Finance, 2019. 19(9): p. 1449-1459.
5. Hu, B., et al. Convolutional neural network architectures for matching natural language sentences. in Advances in neural
information processing systems. 2014.
6. Sundermeyer, M., R. Schlüter, and H. Ney. LSTM neural networks for language modeling. in Thirteenth annual conference
of the international speech communication association. 2012.
7. He, T. and J. Droppo. Exploiting LSTM structure in deep neural networks for speech recognition. in 2016 IEEE Interna-
tional Conference on Acoustics, Speech and Signal Processing (ICASSP). 2016. IEEE.
8. Maqsood, H., et al., A local and global event sentiment based efficient stock exchange forecasting using deep learning.
International Journal of Information Management, 2020. 50: p. 432-451.
9. Patil, P., et al. Stock Market Prediction Using Ensemble of Graph Theory, Machine Learning and Deep Learning Models.
in Proceedings of the 3rd International Conference on Software Engineering and Information Management. 2020.
10. Sim, HS, H.I. Kim, and JJ Ahn, Is deep learning for image recognition applicable to stock market prediction? Complexity,
2019. 2019.
11. Hoseinzade, E. and S. Haratizadeh, CNNpred: CNN-based stock market prediction using a diverse set of variables. Expert
Systems with Applications, 2019. 129: p. 273-285.
12. Eapen, J., D. Bein, and A. Verma. Novel deep learning model with CNN and bi-directional LSTM for improved stock market
index prediction. in 2019 IEEE 9th annual computing and communication workshop and conference (CCWC). 2019. IEEE.
13. Yang, H., Y. Zhu, and Q. Huang. A multi-indicator feature selection for CNN-driven stock index prediction. in International
conference on neural information processing. 2018. Springer.
14. Cai, S., et al. Financial News Quantization and Stock Market Forecast Research Based on CNN and LSTM. in International
Conference on Smart Computing and Communication. 2018. Springer.
15. Oncharoen, P. and P. Vateekul. Deep Learning Using Risk-Reward Function for Stock Market Prediction. in Proceedings
of the 2018 2nd International Conference on Computer Science and Artificial Intelligence. 2018.
16. Liu, Y., et al. Stock price movement prediction from financial news with deep learning and knowledge graph embedding.
in Pacific Rim Knowledge Acquisition Workshop. 2018. Springer.
17. Selvin, S., et al. Stock price prediction using LSTM, RNN and CNN-sliding window model. in 2017 international conference
on advances in computing, communications and informatics (icacci). 2017. IEEE.
18. Liu, S., C. Zhang, and J. Ma. CNN-LSTM neural network model for quantitative strategy analysis in stock markets. in
International Conference on Neural Information Processing. 2017. Springer.
19. Gudelek, M.U., S.A. Boluk, and A.M. Ozbayoglu. A deep learning based stock trading model with 2-D CNN trend detec-
tion. in 2017 IEEE Symposium Series on Computational Intelligence (SSCI). 2017. IEEE.
20. Ding, X., et al. Deep learning for event-driven stock prediction. in Twenty-fourth international joint conference on artificial
intelligence. 2015.
21. Nikou, M., G. Mansourfar, and J. Bagherzadeh, Stock price prediction using DEEP learning algorithm and its comparison
with machine learning algorithms. Intelligent Systems in Accounting, Finance and Management, 2019. 26(4): p. 164-174.
22. Fazeli, A. and S. Houghten. Deep Learning for the Prediction of Stock Market Trends. in 2019 IEEE International Conference
on Big Data (Big Data). 2019. IEEE.
23. Xu, Y. and V. Keselj. Stock Prediction using Deep Learning and Sentiment Analysis. in 2019 IEEE International Conference
on Big Data (Big Data). 2019. IEEE.
24. Lakshminarayanan, S.K. and J. McCrae, A Comparative Study of SVM and LSTM Deep Learning Algorithms for Stock
Market Prediction. 2019.
25. Rana, M., M.M. Uddin, and M.M. Hoque. Effects of Activation Functions and Optimisers on Stock Price Prediction using
LSTM Recurrent Networks. in Proceedings of the 2019 3rd International Conference on Computer Science and Artificial
Intelligence. 2019.
26. Naik, N. and B.R. Mohan. Study of Stock Return Predictions Using Recurrent Neural Networks with LSTM. in International
Conference on Engineering Applications of Neural Networks. 2019. Springer.
27. Nguyen, D.H.D., L.P. Tran, and V. Nguyen. Predicting Stock Prices Using Dynamic LSTM Models. in International Con-
ference on Applied Informatics. 2019. Springer.
28. Lai, C.Y., R.-C. Chen, and R.E. Caraka. Prediction Stock Price Based on Different Index Factors Using LSTM. in 2019 Inter-
national Conference on Machine Learning and Cybernetics (ICMLC). 2019. IEEE.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 32 of 34

29. Hossain, M.A., et al. Hybrid Deep Learning Model for Stock Price Prediction. in 2018 IEEE Symposium Series on Compu-
tational Intelligence (SSCI). 2018. IEEE.
30. Baek, Y. and H.Y. Kim, ModAugNet: A new forecasting framework for stock market index value with an overfitting pre-
vention LSTM module and a prediction LSTM module. Expert Systems with Applications, 2018. 113: p. 457-480.
31. Kim, H.Y. and CH Won, Forecasting the volatility of stock price index: A hybrid model integrating LSTM with multiple
GARCH-type models. Expert Systems with Applications, 2018. 103: p. 25-37.
32. Li, H., Y. Shen, and Y. Zhu. Stock price prediction using attention-based multi-input LSTM. in Asian Conference on Ma-
chine Learning. 2018.
33. Cheng, L.-C., Y.-H. Huang, and M.-E. Wu. Applied attention-based LSTM neural networks in stock prediction. in 2018 IEEE
International Conference on Big Data (Big Data). 2018. IEEE.
34. Shah, D., W. Campbell, and F.H. Zulkernine. A Comparative Study of LSTM and DNN for Stock Market Forecasting. in
2018 IEEE International Conference on Big Data (Big Data). 2018. IEEE.
35. Lin, B.-S., W.-T. Chu, and C.-M. Wang. Application of Stock Analysis Using Deep Learning. in 2018 7th International Con-
gress on Advanced Applied Informatics (IIAI-AAI). 2018. IEEE.
36. Skehin, T., M. Crane, and M. Bezbradica. Day ahead forecasting of FAANG stocks using ARIMA, LSTM networks and
wavelets. 2018. CEUR Workshop Proceedings.
37. Zhang, X. and Y. Tan. Deep Stock Ranker: A LSTM Neural Network Model for Stock Selection. in International Conference
on Data Mining and Big Data. 2018. Springer.
38. Achkar, R., et al. Comparison of BPA-MLP and LSTM-RNN for Stocks Prediction. in 2018 6th International Symposium on
Computational and Business Intelligence (ISCBI). 2018. IEEE.
39. Zeng, Y. and X. Liu. A-Stock Price Fluctuation Forecast Model Based on LSTM. in 2018 14th International Conference on
Semantics, Knowledge and Grids (SKG). 2018. IEEE.
40. Shao, X., et al. Short-term forecast of stock price of multi-branch LSTM based on K-means. in 2017 4th International Con-
ference on Systems and Informatics (ICSAI). 2017. IEEE.
41. Zhao, Z., et al. Time-weighted lstm model with redefined labeling for stock trend prediction. in 2017 ieee 29th international
conference on tools with artificial intelligence (ictai). 2017. IEEE.
42. Nelson, D.M., A.C. Pereira, and RA de Oliveira. Stock market's price movement prediction with LSTM neural networks. in
2017 International joint conference on neural networks (IJCNN). 2017. IEEE.
43. dos Santos Pinheiro, L. and M. Dras. Stock market prediction with deep learning: A character-based neural language model
for event-based trading. in Proceedings of the Australasian Language Technology Association Workshop 2017.
44. Buczkowski, P. Predicting stock trends based on expert recommendations using gru/lstm neural networks. in International
Symposium on Methodologies for Intelligent Systems. 2017. Springer.
45. Chen, K., Y. Zhou, and F. Dai. A LSTM-based method for stock returns prediction: A case study of China stock market. in
2015 IEEE international conference on big data (big data). 2015. IEEE.
46. Song, Y., J.W. Lee, and J. Lee, A study on novel filtering and relationship between input-features and target-vectors in a
deep learning model for stock price prediction. Applied Intelligence, 2019. 49(3): p. 897-911.
47. Naik, N. and B.R. Mohan. Stock Price Movements Classification Using Machine and Deep Learning Techniques-The Case
Study of Indian Stock Market. in International Conference on Engineering Applications of Neural Networks. 2019.
Springer.
48. Chatzis, S.P., et al., Forecasting stock market crisis events using deep and statistical machine learning techniques. Expert
Systems with Applications, 2018. 112: p. 353-371.
49. Abe, M. and H. Nakayama. Deep learning for forecasting stock returns in the cross-section. in Pacific-Asia Conference on
Knowledge Discovery and Data Mining. 2018. Springer.
50. Nakagawa, K., T. Uchida, and T. Aoshima. Deep factor model. in ECML PKDD 2018 Workshops. 2018. Springer.
51. Chong, E., C. Han, and F.C. Park, Deep learning networks for stock market analysis and prediction: Methodology, data
representations, and case studies. Expert Systems with Applications, 2017. 83: p. 187-205.
52. Singh, R. and S. Srivastava, Stock prediction using deep learning. Multimedia Tools and Applications, 2017. 76(18): p.
18569-18584.
53. Ni, L., et al., forecasting of forex time series data based on deep learning. Procedia computer science, 2019. 147: p. 647-652.
54. Li, C., D. Song, and D. Tao. Multi-task recurrent neural networks and higher-order Markov random fields for stock price
movement prediction: Multi-task RNN and higer-order MRFs for stock price classification. in Proceedings of the 25th ACM
SIGKDD International Conference on Knowledge Discovery & Data Mining. 2019.
55. Chen, W., et al., Leveraging social media news to predict stock index movement using RNN-boost. Data & Knowledge
Engineering, 2018. 118: p. 14-24.
56. Zhang, R., Z. Yuan, and X. Shao. A New Combined CNN-RNN Model for Sector Stock Price Analysis. in 2018 IEEE 42nd
Annual Computer Software and Applications Conference (COMPSAC). 2018. IEEE.
57. Li, Y., P. Ni, and V. Chang. An empirical research on the investment strategy of stock market based on deep reinforcement
learning model. in 4th International Conference on Complexity, Future Information Systems and Risk. 2019. SciTePress.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 33 of 34

58. Shin, H.-G., I. Ra, and Y.-H. Choi. A Deep Multimodal Reinforcement Learning System Combined with CNN and LSTM
for Stock Trading. in 2019 International Conference on Information and Communication Technology Convergence (ICTC).
2019. IEEE.
59. Jia, W., et al. Quantitative Trading on Stock Market Based on Deep Reinforcement Learning. in 2019 International Joint
Conference on Neural Networks (IJCNN). 2019. IEEE.
60. Carapuço, J., R. Neves, and N. Horta, Reinforcement learning applied to Forex trading. Applied Soft Computing, 2018. 73:
p. 783-794.
61. Kang, Q., H. Zhou, and Y. Kang. An Asynchronous Advantage Actor-Critic Reinforcement Learning Method for Stock
Selection and Portfolio Management. in Proceedings of the 2nd International Conference on Big Data Research. 2018.
62. Zhu, Y., et al. An Adaptive Box-Normalization Stock Index Trading Strategy Based on Reinforcement Learning. in Interna-
tional Conference on Neural Information Processing. 2018. Springer.
63. Si, W., et al. A multi-objective deep reinforcement learning approach for stock index future's intraday trading. in 2017 10th
International symposium on computational intelligence and design (ISCID). 2017. IEEE.
64. Pastore, A., U. Esposito, and E. Vasilaki. Modelling stock-market investors as reinforcement learning agents. in 2015 IEEE
International Conference on Evolving and Adaptive Intelligent Systems (EAIS). 2015. IEEE.
65. Long, W., Z. Lu, and L. Cui, Deep learning-based feature engineering for stock price movement prediction. Knowledge-
Based Systems, 2019. 164: p. 163-173.
66. Wu, J.-L., et al. A Deep Learning Model for Dimensional ValenceArousal Intensity Prediction in Stock Market. in 2019 IEEE
10th International Conference on Awareness Science and Technology (iCAST). 2019. IEEE.
67. Cho, C.-H., et al. Toward Stock Price Prediction using Deep Learning. in Proceedings of the 12th IEEE/ACM International
Conference on Utility and Cloud Computing Companion. 2019.
68. Minh, D.L., et al., Deep learning approach for short-term stock trends prediction based on two-stream gated recurrent unit
network. IEEE Access, 2018. 6: p. 55392-55404.
69. Hu, G., et al. Deep stock representation learning: from candlestick charts to investment decisions. in 2018 IEEE International
Conference on Acoustics, Speech and Signal Processing (ICASSP). 2018. IEEE.
70. Hu, Z., et al. Listening to chaotic whispers: A deep learning framework for news-oriented stock trend prediction. in Pro-
ceedings of the eleventh ACM international conference on web search and data mining. 2018.
71. Yiqi Zhao and Matloob Khushi. Wavelet Denoised-ResNet CNN and LightGBM Method to Predict Forex Rate of Change,
in 2020 IEEE International Conference on Data Mining Workshops (ICDMW), Sorrento, Italy, 2020
72. Tae Wan Kim and Matloob Khushi. Portfolio Optimization with 2D Relative-Attentional Gated Transformer, in 2020 IEEE
Asia-Pacific Conference on Computer Science and Data Engineering (CSDE), Gold Coast, Australia, 2020
73. Z. Zeng and M. Khushi, "Wavelet Denoising and Attention-based RNN- ARIMA Model to Predict Forex Price," 2020 Inter-
national Joint Conference on Neural Networks (IJCNN), Glasgow, United Kingdom, 2020, pp. 1-7, doi:
10.1109/IJCNN48605.2020.9206832.
74. Z. Zhang and M. Khushi, "GA-MSSR: Genetic Algorithm Maximizing Sharpe and Sterling Ratio Method for RoboTrading,"
2020 International Joint Conference on Neural Networks (IJCNN), Glasgow, United Kingdom, 2020, pp. 1-8, doi:
10.1109/IJCNN48605.2020.9206647.
75. Ling Qi, Matloob Khushi and Josiah Poon. Event-Driven LSTM For Forex Price Prediction, in 2020 IEEE Asia-Pacific Con-
ference on Computer Science and Data Engineering (CSDE), Gold Coast, Australia, 2020Meng, TL; Khushi, M. Reinforce-
ment Learning in Financial Markets. Data 2019, 4, 110.
76. Chen, Sheng, and Hongxiang He. “Stock Prediction Using Convolutional Neural Network.” IOP Conference Series: Mate-
rials Science and Engineering, vol. 435, no. 1, 2018, p. 12026.
77. Gudelek, M. U., Boluk, S. A., & Ozbayoglu, A. M. (2017). A deep learning based stock trading model with 2-D CNN trend
detection. In 2017 IEEE Symposium Series on Computational Intelligence (SSCI) (pp. 1–8).
78. Chen, Jou-Fan, et al. “Financial Time-Series Data Analysis Using Deep Convolutional Neural Networks.” 2016 7th Interna-
tional Conference on Cloud Computing and Big Data (CCBD), 2016, pp. 87–92.
79. Hu, Ziniu, et al. “Listening to Chaotic Whispers: A Deep Learning Framework for News-Oriented Stock Trend Prediction.”
Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining, 2018, pp. 261–269.
80. Pang, Xiong Wen, et al. “An Innovative Neural Network Approach for Stock Market Prediction.” The Journal of Super-
computing, vol. 76, no. 3, 2020, pp. 2098–2118.
81. Feng, Fuli, et al. “Temporal Relational Ranking for Stock Prediction.” ACM Transactions on Information Systems, vol. 37,
no. 2, 2019, pp. 1–30.
82. Chung, Hyejung, and Kyung-shik Shin. “Genetic Algorithm-Optimized Long Short-Term Memory Network for Stock Mar-
ket Prediction.” Sustainability, vol. 10, no. 10, 2018, p. 3765.
83. Li, Jiahong, et al. “Sentiment-Aware Stock Market Prediction: A Deep Learning Method.” 2017 International Conference
on Service Systems and Service Management, 2017, pp. 1–6.
84. Zhang, Kang, et al. “Stock Market Prediction Based on Generative Adversarial Network.” Procedia Computer Science, vol.
147, 2019, pp. 400–406.
Appl. Syst. Innov. 2021, 4, x FOR PEER REVIEW 34 of 34

85. Zhong, X., and David Lee Enke. “Predicting the Daily Return Direction of the Stock Market Using Hybrid Machine Learn-
ing Algorithms.” Financial Innovation, vol. 5, no. 1, 2019, pp. 1–20.
86. Shi, Lei, et al. “DeepClue: Visual Interpretation of Text-Based Deep Stock Prediction.” IEEE Transactions on Knowledge
and Data Engineering, vol. 31, no. 6, 2019, pp. 1094–1108.
87. Yu, Pengfei, and Xuesong Yan. “Stock Price Prediction Based on Deep Neural Networks.” Neural Computing and Appli-
cations, vol. 32, no. 6, 2020, pp. 1609–1628.
88. Souma, Wataru, et al. “Enhanced News Sentiment Analysis Using Deep Learning Methods.” Journal of Computational
Social Science, vol. 2, no. 1, 2019, pp. 33–46.
89. Yong, Bang Xiang, et al. “A Stock Market Trading System Using Deep Neural Network.” Asian Simulation Conference,
2017, pp. 356–364.
90. Wen, Min, et al. “Stock Market Trend Prediction Using High-Order Information of Time Series.” IEEE Access, vol. 7, 2019,
pp. 28299–28308.
91. Jin, Zhigang, et al. “Stock Closing Price Prediction Based on Sentiment Analysis and LSTM.” Neural Computing and Ap-
plications, vol. 32, no. 13, 2020, pp. 9713–9729.
92. Long, Jiawei, et al. “An Integrated Framework of Deep Learning and Knowledge Graph for Prediction of Stock Price Trend:
An Application in Chinese Stock Exchange Market.” Applied Soft Computing, vol. 91, 2020, p. 106205.
93. Chen, Mu-Yen, et al. “Modeling Public Mood and Emotion: Stock Market Trend Prediction with Anticipatory Computing
Approach.” Computers in Human Behavior, vol. 101, 2019, pp. 402–408.
94. Qian, Fei, and Xianfu Chen. “Stock Prediction Based on LSTM under Different Stability.” 2019 IEEE 4th International Con-
ference on Cloud Computing and Big Data Analysis (ICCCBDA), 2019.
95. Li, Zhixi, and Vincent Tam. “Combining the Real-Time Wavelet Denoising and Long-Short-Term-Memory Neural Network
for Predicting Stock Indexes.” 2017 IEEE Symposium Series on Computational Intelligence (SSCI), 2017, pp. 1–8.