International Journal on Recent and Innovation Trends in Computing and Communication

ISSN: 2321-8169 Volume: 11 Issue: 9s

Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

Technical Analysis-Based Data Mining Strategies for

Stock Market Trend Observation
Bhagyashree Pathak
Research Scholar
Department of Computer Science
SAGE University, Indore, (M.P.), India
[email protected]

Dr. Snehlata Barade

Research Supervisor,
Professor and Head of CSE/IT, SAGE University, Indore, (M.P.), India
[email protected]

Abstract— This study introduces a comprehensive approach that utilizes technical analysis-based data mining strategies to
observe and predict stock market trends, by leveraging historical trading data, technical indicators such as moving averages, RSI,
and MACD, to systematically analyze and interpret market behavior, thereby providing investors and traders with actionable
insights for making informed decisions in the volatile environment of stock trading. By integrating quantitative analysis with
predictive modeling, the methodology aims to enhance the accuracy of trend forecasts and identify profitable trading
opportunities. Through the application of cross-validation and backtesting techniques, the effectiveness of these strategies is
rigorously evaluated against actual market movements, offering a robust framework for risk management and portfolio
optimization. This interdisciplinary approach not only demystifies the complexities of the stock market but also opens new
avenues for research and development in financial technology, promising a significant contribution to the field of economic
forecasting and investment strategy.
Keywords- Technical Analysis , Technical Indicators , Stock Market Trend , Data Mining , Neural Network.

I. INTRODUCTION tough times which he may not stand. Any pattern is well
backtested and thus to be followed as it is.
1.1 Overview 1.2 Introduction of Data Mining
This is my first research as a student so I decided to pick a
topic which is center of attraction for many people and is of
In this thesis we are using Neural Network as the application
wide use also. In this process, I came across many topics but
of Data Mining concept for predicting the trend in stock
as I have to use data mining techniques on my related domain,
market. Data mining is a process of discovering various
I chose stock market as a co-domain because of increasing
models, summaries, and derived values from a given
amount of financial data available on internet. Also it is widely
collection of data. It is one of the fastest growing fields in the
discussed among people and everybody is somehow affected
computer industry and provides a wide range of
and concerned about stock market’s direction as they are
methodologies and techniques that can be applied to a host of
linked with it directly or indirectly.
problem sets [5][6]. Many data mining techniques have been
As I liked the topic, I tried to find out more advantages of this
applied to stock data to achieve better financial solutions
selection. I came across with the following benefits.
[1,2,3,4 ]
1. It can be good source of secondary income.
Data mining refers to the finding of relevant and useful
2. It is very randomly traded by common man.
information from databases. Data mining and knowledge
3. 85 % of people lose money in stock market in long run by
discovery is a new interdisciplinary field, merging ideas from
trading in different instruments.
statistics, machine learning, data bases and parallel computing.
4. It is a big center of attraction for the people especially who
Researchers have defined the term ‘data mining’ in many
are the earning members of their family.
ways: [39][39]
5. Stock market is seen as quick money generator.
Data mining or knowledge discovery in data bases, as it is also
This thesis focuses on generation of trading signals with the known, is the non trivial extraction of implicit, previously
help of technical analysis. The main feature of any algorithm unknown and potentially useful information from the data.
[7] [30]
based trading system is that it eliminates the weaknesses and The amount of data continues to grow at an enormous rate
improper interference of a trader, and guides it through the even though the data stores are already vast. [31] The primary
challenge is how to make the database a competitive business

1050
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

advantage by converting seemingly meaningless data into Richard et al. [33] says that, Data mining is the process of
useful information. [32] How this challenge is met is critical employing one or more computer learning techniques to
because companies are increasingly relying on effective automatically analyze and extract data.
analysis of the information simply to remain competitive. A Edelstein et al. [34], states that data mining involves the use of
mixture of new techniques and technology is emerging to help sophisticated data analysis tools to discover previously
sort through the data and find useful competitive data.[8][9] unknown, valid patterns and relationships in large data sets.
[29] By knowledge discovery in databases, interesting Nitchi et al. [35] gives a simple definition considers the data
knowledge, regularities, or high-level information can be mining as the process of extracting predictive information
extracted from the relevant sets of data in databases and be hidden in large datasets.
investigated from different angles, [40] and large databases Mining information and knowledge from large database has
thereby serve as rich and reliable sources for knowledge been recognized by many researchers as a key research topic in
generation and verification. database systems and machine learning. Companies in Many
industries also take knowledge discovering as an important area
1.3 Basics
with an opportunity of major revenue (Fayyad et al., 1996.
Piatetsky-Shapiro et al. 1991, Silberschatz ET al.1995). The
Two main parts of financial analysis are: discovered knowledge can be applied to information
1. Fundamental analysis management, query processing, decision making, process
2. Technical analysis control, and many other applications.
2.2 Neural Network
1.3.1 Fundamental Analysis:
It’s a relatively new process of data mining in computing
This is done specially by professional analysts who have taken which is based on development of a system through the ability
degree in commerce field such as CA, CS, MBA in finance to learn and recognize patterns in a given set of data, which is
etc. This analysis based on financial data provided by very complex for any computer techniques to find and
company every quarter, which includes financial attributes establish. Several elements are interconnected like neurons of
such as growth forecast, PE ratio, dividend yield, upper and brain which then provide a pattern which is useful in some
lower base line estimations, balance sheets etc. It also depends sense.
on the quality of management running the company. It needs Berry et al. [36], according to neural networks are the most
updating very frequently and detailed analysis is required. widely used known and the least understood of the major data
[10][11] [28] mining techniques. He says training a neural network is a
1.3.2 Technical analysis: process that involves setting weights on inputs to best
It is focused on past history of share performance, analyzed approximate a target variable. This is important for optimizing
statistically through various formulas and indicators such as the neural network. Three steps are involved in training.
moving averages (MA). It emphasize on predicting the trend Training instance variables, calculating outputs using existing
of stock and make use of competitive trading among institutes. weights and calculating errors and accordingly adjusting
weights.
II. LITERATURE REVIEW Pyle et al. [37], describes neural networks as network
construction network system of interconnected interacting
In previous chapter we discussed the basic concept of data weights at each node acting as input and output stations. Each
mining. It is a general definition of data mining. Now a day’s input to the network gets its own node which consists of a
data mining has been extensively used in many fields, Such as transformation of input variables fed in. The input unit is
banking and financial markets. In banking industry it is connected to the output unit with a weighting and the input is
heavily used to model and predict credit fraud, to evaluate combined in the output unit with a combination function. The
risk, to perform trend analysis, and to analyze profitability, as activation function is the passed transfer function.
well as to help with direct marketing campaigns. [12]
In the financial markets, neural networks have been used in Singh et al., 2022, This study aims to advance and introduce a
stock markets to analyze the trend, stock price forecasting, in more rigorous neural network approach for predicting stock
option trading, in bond rating, in portfolio management, in market trends. The present study presents a comprehensive
commodity price prediction, in mergers and acquisitions, as examination of stock market forecasting using neural
well as in forecasting financial disasters. [12] networks. The analysis of stock market forecast may be
In this chapter we present small introduction on data mining conducted using a wide range of machine learning techniques.
and few research papers discussion on our subject. Most This study investigates the buying and selling prices of stocks
papers are based on application of neural network in stock on the Bombay Stock Exchange (BSE) and analyses the
market. provided data. To predict stock market trends, I used a
comprehensive Long-Short-Term Memory Neural Network
2.1 Definitions of Data mining by different Authors
(LSTM) with an embedded layer and LSTM Neural Network.
Applying deep LSTM with layers improves the experimental

1051
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

layout. The prediction technology demonstrates precise Absolute Percentage Error (MAPE) of at least 10%. Among
outcomes that provide exceptional financial gains. [41] all the investigated models, Model 6-20-20-1 stands out as the
Liu et al., 2022, Forecasting stock market indices is an most superior, with an MSE of 0.00055, RMSE of 0.007418,
alluring subject. Conducting a thorough examination of this and MAPE of 1.17%. [43]
subject matter will provide significant insights for investors, Sharma et al., 2022, Stock market forecasting often use both
traders, and policymakers involved in the attractive stock traditional statistical methods and artificial intelligence tools.
markets. This paper presents a novel sparrow search technique The presence of nonlinearity in stock data might lead to
that aims to enhance the initial weight and threshold inaccurate forecasting results when employing a conventional
estimation of BPNN. The algorithm is specifically designed or single intelligent approach. Hence, it is essential to cultivate
for the purpose of predicting stock market indices. This article a fusion of intelligent methodologies in order to construct a
primarily focuses on the following aspects: (1) The sine chaos proficient prediction model. This article presents a novel
model is used for population initialization. (2) The position approach to forecasting by combining an Artificial Neural
update formula of the discoverer incorporates the global Network (ANN) with a Genetic Algorithm (GA). The
optimum solution from the preceding generation, together with proposed technique utilises two prominent US stock market
adaptive weights to optimise the coordination between local indices, namely DOW30 and NASDAQ100, for the purpose of
mining and global exploration capabilities. In order to develop forecasting. The datasets were divided into three subsets:
novel solutions, the Gaussian mutation operator and the training, testing, and validation. The validation of the model
reverse learning technique are used to perform perturbation was conducted using the stock data during the COVID-19
mutation at the ideal point. In order to optimise the initial period. The experimental results obtained using the DOW30
weight and threshold of the BP neural network, the improved and NASDAQ100 datasets demonstrate that the GA and ANN
sparrow algorithm (ISSA) is used. The suggested model's hybrid model outperforms the single ANN (BPANN) method
performance is assessed on four datasets, namely SSE, SZSE, in terms of accuracy for both the DOW30 and NASDAQ100,
SP500, and DJI. The suggested mode is contrasted with two both in the short and long term. [44]
kinds of models. One approach involves using swarm
intelligence optimisation algorithms, such as GA-BP, PSO-BP, Srivinay et al., 2022, The volatility of stock prices may be
ACO-BP, GWO-BP, CS-BP, and SSA-BP, to optimise BPNN. attributed to several elements inherent in the stock market,
The remaining models are classified as deep learning models. including geopolitical tensions, corporate profits, and
The empirical findings demonstrate that the three solutions put commodity prices, all of which have an influence on stock
forward in this study significantly enhance the optimisation prices. Certain factors, such as reserve bank policy,
efficacy of the sparrow search algorithm. The ISSA-BP model government policy, inflation, and global market uncertainty,
has shown significant efficacy in the realm of short-term stock might influence stock prices. Estimating the volatility of
price prediction, therefore enabling investors to forecast stocks is a significant challenge for traders. A precise forecast
market trends and identify optimal trading opportunities. of stock prices aids investors in mitigating the risk associated
Furthermore, policymakers may evaluate the rationale of with their portfolios or investments. Equity prices exhibit
policies by analysing market patterns, so enhancing the nonlinearity. We suggest using a hybrid stock prediction
advancement of the stock market. [42] model that combines the prediction rule ensembles (PRE)
approach with deep neural network (DNN) to address the issue
Napitupulu et al., 2022, The worldwide COVID-19 outbreak of nonlinearity in the data. The identification of an upward
has elicited widespread anxiety. Furthermore, it caused trend in stock prices is facilitated by the use of stock technical
significant disruptions to global life and economic operations. indicators. The moving average technical indicators that were
Forecasting the stock market throughout the COVID-19 taken into consideration are the moving average of 20 days,
epidemic was a significant obstacle because to the non- the moving average of 50 days, and the moving average of 200
stationary, unpredictable, and intricate nonlinear nature of the days. Furthermore, the PRE approach was used to calculate
data. Therefore, it is essential to conduct a comprehensive many rules for stock prediction. Subsequently, the rules
examination of the aforementioned patterns in order to exhibiting the lowest root mean square error (RMSE) score
construct a suitable prognostic framework for forecasting the were chosen. Furthermore, the three-layer deep neural network
stock market within the pandemic. The objective of this (DNN) is being used for stock prediction. The
project is to develop a stock market prediction model for the hyperparameters of the deep neural network (DNN) have been
Indonesia Stock Exchange during the COVID-19 epidemic. optimised, including the number of layers, learning rate,
The model uses a deep learning method that relies on artificial number of neurons, and number of epochs in the model.
neural networks. This study focuses on the pharmaceutical Furthermore, the PRE and DNN prediction models are
business within the health sector that is publicly traded on the aggregated to get the average results. The outcomes of the
IDX. The factors being considered in this study include the hybrid stock prediction model are calculated using the mean
suggested model for forecasting stock prices, which absolute error (MAE) and root mean square error (RMSE)
incorporates daily stock price fluctuations, as well as COVID- metrics. The hybrid stock prediction model has superior
19 trend indicators and the government's reaction tightness performance compared to the individual prediction models,
index to COVID-19 in Indonesia. The research findings namely the DNN and ANN, exhibiting a 5% to 7%
indicate that all suggested model systems have exceptional enhancement in the root mean square error (RMSE) score. The
accuracy in predicting stock market prices, with a Mean study takes into account the Indian stock price data. [45]

1052
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

financial market risks and maximising investment returns.

Manujakshi et al., 2022, The volatility of stock prices may be This subject has garnered substantial interest from both
attributed to several elements inherent in the stock market, academic and business spheres. Nevertheless, due to its
including geopolitical tensions, corporate profits, and dynamic and intricate nature, the stock market is subject to the
commodity prices, all of which have an influence on stock influence of many causes and has a complete capacity to
prices. Certain factors, such as reserve bank policy, incorporate intricate financial information. The analysis and
government policy, inflation, and global market uncertainty, processing of multi-source heterogeneous data in the stock
might influence stock prices. Estimating the volatility of market pose significant constraints for present intelligent
stocks is a significant challenge for traders. A precise forecast algorithms due to the various, heterogeneous, and complicated
of stock prices aids investors in mitigating the risk associated nature of the explanatory variables of influencing factors.
with their portfolios or investments. Equity prices exhibit Hence, this research employs the edge weight and information
nonlinearity. We suggest using a hybrid stock prediction transmission mechanism that is well-suited for subgraph data
model that combines the prediction rule ensembles (PRE) in order to thoroughly screen nodes. Additionally, the gate
approach with deep neural network (DNN) to address the issue recurrent unit (GRU) and long short-term memory (LSTM) are
of nonlinearity in the data. The identification of an upward used to combine subgraph nodes. The collected data includes
trend in stock prices is facilitated by the use of stock technical the metapaths of three categories of index data, and the
indicators. The moving average technical indicators that were incorporation of the association relationship attention
taken into consideration are the moving average of 20 days, dimension efficiently extracts the underlying meanings of
the moving average of 50 days, and the moving average of 200 diverse data from several sources. The integration of the
days. Furthermore, the PRE approach was used to calculate metapath attention mechanism with a graph neural network
many rules for stock prediction. Subsequently, the rules enables the comprehensive categorization of diverse graph
exhibiting the lowest root mean square error (RMSE) score data from several sources. This integration facilitates the
were chosen. Furthermore, the three-layer deep neural network prediction of stock market volatility. The findings indicate that
(DNN) is being used for stock prediction. The the aforementioned approach is viable for integrating diverse
hyperparameters of the deep neural network (DNN) have been stock market data and extracting implicit semantic information
optimised, including the number of layers, learning rate, related to association relationships. The suggested technique in
number of neurons, and number of epochs in the model. this research demonstrates a 16.64% increase in accuracy
Furthermore, the PRE and DNN prediction models are compared to the dimensional reduction index and a 14.48%
aggregated to get the average results. The outcomes of the increase compared to other methods used for fusion and
hybrid stock prediction model are calculated using the mean prediction of heterogeneous data using the same model. [48]
absolute error (MAE) and root mean square error (RMSE)
metrics. The hybrid stock prediction model has superior Li and Qian, 2022, Forecasting stock prices is essential but
performance compared to the individual prediction models, also difficult in any stock market trading strategy. Presently,
namely the DNN and ANN, exhibiting a 5% to 7% the use of recurrent neural networks (RNNs) has become
enhancement in the root mean square error (RMSE) score. The prevalent in the field of stock prediction, yielding several
study takes into account the Indian stock price data. [46] notable achievements. Nevertheless, challenges persist in
enhancing the efficacy of RNNs in a complex stock market.
Sako et al., 2022, The forecasting of financial and economic
RNNs are insufficient in extracting distinctive characteristics
time series has historically posed challenges because to its
susceptibility to political, economic, and social influences. from a multitude of signals in the flow of stock information.
Due to this rationale, individuals engaged in financial market Furthermore, while using RNN, a solitary long time cell from
investments and currency exchange often seek resilient models the market is often combined into a solitary feature, resulting
that can effectively optimise their portfolio and mitigate in the loss of crucial temporal information necessary for
potential losses. Recent research have shown that Recurrent accurate stock prediction. In this study, we propose a new
Neural Networks (RNNs), a specific form of Artificial Neural hybrid neural network called the frequency decomposition
Networks (ANNs), might enhance the prediction accuracy of induced gate recurrent unit (GRU) transformer, also known as
financial data over time. The objective of this study is to FDGRU-transformer or FDG-trans, to address these two
predict the closing prices of eight stock market indexes and six problems. In FDG-transformer, we use empirical model
foreign exchange rates associated with the USD. This will be decomposition to break down the whole ensemble of cluttered
achieved by using the Recurrent Neural Networks (RNNs)
data into a trend component and many informative and
model, namely the Long Short-Term Memory (LSTM) and the
independent mode components, drawing inspiration from the
Gated Recurrent Unit (GRU) variations. The findings indicate
that the GRU model yields superior outcomes, particularly in success of frequency decomposition. With the use of
the context of univariate out-of-sample forecasting for decomposition, the FDG-transformer has the capability to
currency exchange rates and multivariate out-of-sample extract distinctive insights from signals that are crowded. The
forecasting for stock market indexes. [47] FDG-transformer employs a hybrid neural network consisting
of GRU, long short term memory (LSTM), and multi-head
Li et al., 2022, The examination of forecasting stock market attention (MHA) transformers to preserve the temporal
volatility has considerable importance in effectively managing information in the observed jumbled data. The integrated

1053
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

transformer network may encode the influence of various

weights from previous time steps to the present one, leading to
the creation of a higher-level time series model. We use the
created FDG-transformer model to analyse Limit Order Book
data and conduct a comparative analysis with the outcomes
acquired from other contemporary methodologies. The
comparative analysis demonstrates that our algorithm
efficiently provides accurate price forecasts. Furthermore, a
comprehensive ablation investigation is undertaken to
substantiate the significance and indispensability of every
constituent inside the suggested model. [49]
Figure 1. Multi-Agent System Architecture
2.3 Previous Work in the field
It also consists of a future signal prediction for X % change in
Nhamo Mdzingw et al, [30] focused on data or turning data a stock price so to make predictions easy. Many efforts are
into information by a process requiring a unique combination made to combine fundamental and technical analysis and this
of tools for each application. He tries to adhoc methodology MASA provides a good model which enhances all stock
nearly used data mining in the commercial world mainly related data and information at one go for an investor to
focusing and the data mining process and data mining outperform stock market through choosing active and
algorithms used. It will also include a brief description of the directional stocks with proper exits.
Oracle data mining tool. K. Senthamarai Kannan et al [22], on the other hand only took
technical analysis as the base of their BSRCTB algorithm and
Boris kovalerchuck et al, [31] describes data mining in finance connected all different statistical indicators to outperform the
by discussing financial tasks, specifies methods and stock and simple individual indicator returns. In BSRCTB
techniques in data mining field. in the first part of thesis they algorithm, they combined Typical Price (TP), Bollinger
includes time dependence, data selection ,forecast horizon, Bands, Relative Strength Index (RSI), CMI and Moving
measures of success, quality of patterns hypothesis evaluation, Average (MA). TP is calculated by averaging the previous
problem id, method profile, attribute based and relational day’s open, high and low. [19] This algorithm beats the
methodologies. In the second part of the thesis discusses data markets n absolute terms and according to authors, it can be
mining modals and practice in finance. It covers use of neural used for trading analysis and is good for swing traders.
networks in use of portfolio management, design of Savinderjit Kaur et al [23], in their article on application of
interpretable trading and discovering money laundering data mining conclude that there is a need to more automation
schema using decision rules and relational data mining of systems and new hybrid systems should be worked on.
methods. They also states that Data mining has been used in stock
Selv Nhamo Mdzingw et al, an Simon et al [20], suggested an market to make predictions regarding trends and prices to
ANN (artificial neural network) model for the prediction in gain maximum profits. A lot of research has been carried out
stock market. They have used the basic 3 steps of neural on its various aspects. [15] There is still room for improving
network viz input layer, middle or hidden layer and final or accuracy of these prediction methods by developing new
output layer. [14] The middle layer consists of many statistical hybrid methods and by improving the existing algorithms.
methods and formulas. It’s hidden because it processes the More research can be done in detecting trends in the stock
information passed from the 1st layer in many steps and result market like studying the abnormal stock returns, trends
is shown to user. Authors also compared many different preceding and following executive stock options awards,
techniques for stock market prediction like RNN, MLP,BNPP book-to-market effect, bubble diagnosis, inter industry
etc and emphasized that better results can be obtained when patterns etc.
these techniques are used with association with their ANN Author, focused on effectiveness of technical analysis and
model for prediction of stock market. Dow’s theory, for this they use the daily data of three stock
Monica Tirea et al, [21] proposed a Multi-Agent Stock market: DJIA, HIS, Taiwan stock market for the technical
Trading Algorithm Model which uses both fundamental and analysis. In this paper mostly they are using buy and hold
technical analysis simultaneously. It basically works for trading strategy. [16] In this paper he proposed a DBN model
investors and not trades who trade frequently. to predict Taiwan stock market. [24]
Robert P. Schumacher et al [25], experimented using several
linguistic textual representations, including Bag of Words,
Noun Phrases, and Named Entities approaches. He believe
that using textual representations other than the de facto
standard Bag of Words will yield improved predictability
results. [17] They also used many Stock Market prediction
taxonomy of the various machine learning techniques such as

1054
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

genetic algorithm, Naïve Bayesian technique, SVM (support measure of volatility, Bollinger Bands adjust themselves to the
vector machine). market conditions. When the markets become more volatile,
S. Abdulsalam Suleiman et al, [26] presented regression the bands widen (move further away from the average), and
analysis as a data mining technique and developed tool for during less volatile periods, the bands contract (move closer to
exploiting especially time series data in financial institution. A the average). The tightening of the bands is often used by
prediction system has been built that uses data mining technical traders as an early indication that the volatility is
technique to produce periodically forecasts about stock market about to increase sharply. The closer the prices move to the
prices his technique complement proven numeric forecasting upper band, the more overbought the market, and the closer
method using regression analysis with technology taking as the prices move to the lower band, the more oversold the
input the financial information. market.
Ayo Charles K. et al [27], identifies that stock market
prediction is one of the most important issue now a days. so In 3.2 Stochastic
this paper, author present a hybridized approach which
combines the use of the variables of technical and fundamental A technical momentum indicator that compares a security's
analysis of stock market indicators for prediction of future closing price to its price range over a given time period. The
price of stock in order to improve on the existing approaches oscillator's sensitivity to market movements can be reduced by
with the help of Ann technique. The hybridized approach was adjusting the time period or by taking a moving average of the
tested with published stock data and the results obtained result. This indicator is calculated with the following formula:
showed remarkable improvement over the use of only %K = 100[(C - L14)/(H14 - L14)]
technical analysis variables. [18] Also, the prediction from
hybridized approach was found satisfactorily adequate as a C = the most recent closing price
guide for traders and investors in making qualitative decisions. L14 = the low of the 14 previous trading sessions
H14 = the highest price traded during the same 14-day period.
III. TECHNICAL PARAMETER STUDIED
%D = 3-period moving average of %K

Our system is totally based on analysis of technical parameters

and there are vast numbers of technical parameters available
with different properties. We checked following technical
indicators with both preceding and lagging properties and
oscillators but considered only those which gave proper results
with less number of whipsaws and stoploss limits and got in
tune with swing trading system. Here, we are giving record of
those technical parameters which e studied but not used.
The most widely used technical indicators are as follows.
3.1 Bollinger Band

A band plotted two standard deviations away from a simple

moving average, developed by famous technical trader John Figure 3. Stochastic
Bollinger.
The theory behind this indicator is that in an upward-trending
market, prices tend to close near their high, and during a
downward-trending market, prices tend to close near their low.
Transaction signals occur when the %K crosses through a
three-period moving average called the"%D".
1) 3.3 RSI (Relative strength index)
A technical momentum indicator that compares the magnitude
of recent gains to recent losses in an attempt to determine
overbought and oversold conditions of an asset. It is calculated
using the following formula:

RSI = 100 - 100/ (1 + RS*)

*Where RS = Average of x days' up closes / Average of x

Figure 2. Bollinger Bands days' down closes.
In this example of Bollinger Bands, the price of the stock is
banded by an upper and lower band along with a 21-day
simple moving average. Because standard deviation is a

1055
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

Figure 4 . RSI (Relative strength index)

As shown in the chart, the RSI ranges from 0 to 100. An asset .

is deemed to be overbought once the RSI approaches the 70
level, meaning that it may be getting overvalued and is a good Figure 6. Parabolic SAR ( Stop and Reverse )
candidate for a pullback. Likewise, if the RSI approaches 30, it
is an indication that the asset may be getting oversold and This method was developed by J. Wells Wilder. Basically, if
therefore likely to become undervalued. the stock is trading below the parabolic SAR (PSAR) you
A trader using RSI should be aware that large surges should sell. If the stock price is above the SAR then you
and drops in the price of an asset will affect the RSI by should buy (or stay long).
creating false buy or sell signals. The RSI is best used as a
valuable complement to other stock-picking tools. 3.6 William’s %R

3.4 ADX (Average Directional Index) In technical analysis, this is a momentum indicator measuring
overbought and oversold levels, similar to a stochastic
An indicator used in technical analysis as an objective value oscillator. It was developed by Larry Williams and compares a
for the strength of trend. ADX is non-directional so it will stock's close to the high-low range over a certain period of
quantify a trend's strength regardless of whether it is up or time, usually 14 days.
down. ADX is usually plotted in a chart window along with
two lines known as the DMI (Directional Movement
Indicators). ADX is derived from the relationship of the DMI
lines. Analysis of ADX is a method of evaluating trend and
can help traders to choose the strongest trends and also how to
let profits run when the trend is strong.

Figure 7. William’s %R

It is used to determine market entry and exit points. The

Williams %R produces values from 0 to -100, a reading over
80 usually indicates a stock is oversold, while readings below
20 suggests a stock is overbought.
Figure 5. ADX
3.7 Candle Stick
3.5. SAR (Stop-and-Reverse)
The Japanese began using technical analysis to trade rice in
A technical analysis strategy that uses a trailing stop and the 17th century. While this early version of technical analysis
reverse method called "SAR," or stop-and-reversal, to was different from the US version initiated by Charles
determine good exit and entry points. Dow around 1900, many of the guiding principles were very
similar:

1056
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

1. The "what" (price action) is more important than the "why"

(news, earnings, and so on). 3.8 Trend Lines
2. All known information is reflected in the price.
3. Buyers and sellers move markets based on expectations Technical analysis is built on the assumption that prices trend.
and emotions (fear and greed). Trend Lines are an important tool in technical analysis for
4. Markets fluctuate. both trend identification and confirmation. A trend line is a
5. The actual price may not reflect the underlying value. straight line that connects two or more price points and then
extends into the future to act as a line of support or resistance.
Many of the principles applicable to support and resistance
levels can be applied to trend lines as well.

Figure 8. Candlestick Formation

Candlestick charting first appeared sometime after 1850.

Figure 10. Trend Line
Much of the credit for candlestick development and charting
goes to a legendary rice trader named Homma from the town Trend lines can offer great insight, but if used improperly, they
of Sakata. It is likely that his original ideas were modified and can also produce false signals. Other items - such as horizontal
refined over many years of trading eventually resulting in the support and resistance levels or peak-and-trough analysis -
system of candlestick charting that we use today. should be employed to validate trend line breaks. While trend
lines have become a very popular aspect of technical analysis,
they are merely one tool for establishing, analyzing, and
confirming a trend. Trend lines should not be the final arbiter,
but should serve merely as a warning that a change in trend
may be imminent. By using trend line breaks for warnings,
investors and traders can pay closer attention to other
confirming signals for a potential change in trend.

IV. QUANTITATIVE TRADING

Quantitative trading is methodology employing advanced

statistical techniques to make trading decision which can be
traded either manually or electronically. With advancement in
computing power, it is advantageous to implement such back-
tested strategies as algorithmic trading which removes chances
of human error significantly. The frequency of trade can be
Figure 9. Candlestick Pattern - Morning Doji Star high or low as per the strategy.
Quantitative investing represents an investing technique
Here given an example of a famous bullish pattern called as typically employed by the most sophisticated, technically
Morning Doji Star, which consists of 3 candles. This pattern is advanced hedge funds. These quant shops employ fast
an indication of a trend reversal from bearish to bullish. The computers to find predictable patterns within financial data.
opposite of this pattern is called as Evening Doji Star which 4.1 Overview on Algorithmic Trading
acts as a trend reversal from bullish to bearish.
Here are some examples of famous patterns of candlesticks:
Automated or Algorithmic trading is using computers to
Bullish Engulfing, Bearish Engulfing, Bullish Harami, Bearish
generate trading signals, sending orders and managing
Harami, Hammer, Hanging Man, Inverted Hammer, Shooting
portfolio using algorithms with or without human initiation.
Star, Morning Star, Piercing Pattern, Dark Cloud Cover, Three
Sophisticated electronic markets/platforms are used by the
White Soldiers, Three Black Crows etc.
algorithms to trade in the similar fashion as done in electronic

1057
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

trading. The difference is that in algorithmic trading decisions • In January 1989, BSE National Index was introduced,
about volume or size, timing and price are determined by the which was renamed as BSE-100 Index from October 14,
algorithm. Furthermore, algorithmic trading efficiently 1996 and launched its dollar-linked version on May 22,
increases the universe being traded by an individual trader 2006.
which is limited in electronic trading environment. • On July 25, 1990, the SENSEX touched the four-digit
figure for the first time and closed at 1,001.
Automated or algorithmic trading is a system for trading • SEBI was formed officially by the Government of India in
through programmed software that can make around 800 1992 with SEBI Act 1992 being passed by the Parliament
transactions per second. The software-helped trading was of India.
allowed by Bombay Stock Exchange (BSE) and National • On 27 May 1994, BSE launched two new index series: The
Stock Exchange (NSE) in 2005. 'BSE-200' and the 'DOLLEX-200'.
A third of all European Union and United States stock trades • In 1995, the BSE switched to an automated, screen-based
in 2006 were driven by automatic programs, or algorithms, trading platform called BSE On-line trading (BOLT),
according to Boston-based financial services industry research which at present has a capacity of 8 million orders per day.
and consulting firm Aite Group. As of 2009, HFT firms In 1997, the BOLT system expanded nation-wide.
account for 73% of all US equity trading volume. • BSE-500 Index and 5 sectorial indices were launched in
1999.
High-Frequency Trading (HFT): • In 2001, BSE launched BSE-PSU Index, DOLLEX-30 and
It is a special category of algorithmic trading characterized by the country's first free-float based index - the BSE TECk
unusually brief position-holding periods, low-latency response Index.
times, and high trading volumes in a day. Algorithms are • The index calculation for the SENSEX was shifted from
written so as to exploit trading opportunities which appear in the ‘full market capitalization’ method to the ‘free float
very brief time periods as short as milli- or micro- seconds. method’ on Sep 1, 2003. Over the years, BSE shifted all its
The margin of each trade is small, which is compensated by indices to the free-float methodology (except BSE-PSU
fast speed and large volumes. index).
4.2 Stock Market History • On June 20, 2005, the news of the settlement between the
Ambani Brothers helped the SENSEX crossed 7,000 points
A Brief History of Indian Stock Market (BSE and NSE) : for the first time.
• On 17th May 2004, second biggest fall of all time, Circuit
In 1957, it became the first stock exchange to be recognized filters used twice in a day (564.71 points, 11.14%).
by the Indian Government under the Securities Contracts • In 2005, the BSE (Corporatization and Demutualization)
Regulation Act. Scheme, 2005 was introduced by SEBI and the exchange
It is the oldest stock exchange in Asia (established in 1875) turned into a corporate entity renamed as Bombay Stock
and is currently located on Dalal Street, Mumbai. Exchange Limited from Bombay Stock Exchange.
• On February 7, 2006, the SENSEX closed above the
• It is the 6th largest stock exchange in Asia and the 14th 10,000-mark.
largest in the world with equity market capitalization of • On March 7, 2007, Singapore Exchange Limited entered
US$1 trillion as of December 2011. into an agreement to invest in a 5% stake in BSE Limited.
• It has also introduced the world's first centralized • Due to effects of the Subprime crisis in the U.S and heavy
exchange-based internet trading system, BSEWEBx.co.in selling in the international markets, the BSE SENSEX fell
to enable investors anywhere in the world to trade on the by 615 points in a single day on Wednesday August 1,
BSE platform. 2007.
• There are over 5,112 listed Indian companies and over • On Jan 08, 2008, The SENSEX touched all time peak of
8,196 scripts on this exchange as of December 2011. 21078 before closing at 20873.
• The World Council of Corporate Governance has awarded
Bombay Stock Exchange (BSE) - Introduction, Facts & the Golden Peacock Global CSR Award for BSE's
Major Evolutions : initiatives in Corporate Social Responsibility (CSR).
• BSE launched SENSEX (BSE 30), a free-float market • On May 18, 2009, the SENSEX surged 2110.79 points
Capitalization-weighted stock market index on January 2, from the previous closing. This event created history in
1986, with a base value 100 and base year 1978-1979. Dalal Street, by being the first ever time that trade had
• The SENSEX consists of 30 well-established and been suspended for an increase in value. This rally is
financially sound companies listed on BSE Limited. These primarily due to the victory of the UPA in the 15th General
Companies represents various industrial sectors of the elections.
Indian economy.
• The market capitalization of the SENSEX was about Rs.
29,275 billion while its free-float market capitalization was
Rs.14,660 billion as of April,2012.

1058
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

National Stock Exchange Limited (NSE) - Introduction, 2 Swing traders hold stocks for days or weeks playing the
Facts & Major Evolutions : general upward or downward trends.
3 Swing Trading is not high-speed day trading. Some
• It was incorporated in November 1992 as a tax-paying people call it momentum investing, because you only
company and in April 1993, it was recognized as a stock hold positions that are making major moves.
exchange under the Securities Contracts Regulation Act, 4 By rolling your money over rapidly through short term
1956. gains you can quickly build up your equity.
• It is the 16th largest stock exchange in the world and
largest in India by daily turnover and number of trades, for 4.5 Proposed Methodology
both equities and derivative trading.
• As of December 2011, it has a market capitalization of As said above, this system tries to be simple as well as
around US$985 billion. effective, so common and simple technical indicators are used
• It is the second fastest growing stock exchange in the here from a wide variety. We studied EMA, their crossovers,
world with a recorded growth of 16.6% and over 1,640 RSI, MACD, ADX, SAR, Candlestick Patterns, Trend Lines
listings as of December 2011. etc. After working on various combinations we tried to find
• It is the third largest Stock Exchange in the world in terms out the main link which provides maximum coverage of
of the number of trades in equities. conditions occurring in stock market. The few main aspects,
• In 1994, Wholesale Debt Market segment Capital Market we found necessary for a trading strategy are as follows:
(Equities) segment goes live. 1. Selection of Time Frame.
• In 1995, the NSE established Investor Grievance Cell, 2. Indicator Spotting Right Market Direction.
NSCCL (the first Clearing Corporation), and Investor 3. Strength in Signal.
Protection Fund. 4. Using the Signals to their Full Capacity.
• In 1996, the NSE launched S&P CNX Nifty (Nifty 50), a 5. Removing Low Probability Trades.
stock market index owned and managed by India Index
4.5.1 Selection of Time Frame
Services and Products Nifty 50 is the largest single
financial product in India, with an ecosystem comprising:
We tried different time frames in our analysis which were,
exchange traded funds (onshore and offshore), exchange-
hourly, 3 hourly, daily and weekly. Hourly were too volatile
traded futures and options, other index funds and OTC
and weekly was untradeable in our selected indicators. So after
derivatives (mostly offshore).
selecting daily close as time interval for our trades, we
• From June 26, 2009, the computation for calculating Nifty optimized other indicators in accordance to it.
Index was changed to free float methodology from full
market capitalization methodology.
4.5.2 Indicator Spotting Right Market Direction
• The base period for the S&P CNX Nifty index is
November 3, 1995. The base value of the index has been
On daily basis the use of ADX indicator helped in a lot to
set at 1000, and a base capital of Rs 2.06 trillion.
decide market direction in near term. It also showed the
• In December 1996, CNX Nifty Junior was launched. It maturity of trade position and also kept the whipsaws to
consists of 50 companies representing approximately 10% minimum ADX have 3 lines viz positive green line, negative
of the traded value of all stocks on the NSE. The CNX red line and calculate ADX third line.
Nifty Junior is owned and operated by India Index Services When green line is above red and ADX is above 20, it’s an
and Products Ltd. indication of start of a positive trade. If red is above green line
• In June 2005, Futures & options in BANK Nifty Index and ADX is above 20 it’s a short trade and markets/stocks are
were launched on National Stock Exchange. weak. Below ADX less than 20, there is no direction present in
• In 2007, the NSE launched derivatives on Nifty Junior & market and hence they should be avoided. ADX considers 14
CNX 100 and derivatives on Nifty Midcap 50. days as it defaults value.
• In August 2009, Interest Rate Futures was launched on this In our trading system, we optimized it and changed it from 14
Exchange. to 8 days and it enhanced the results. As we reduced the
• The NSE controls more than 90 percent of India’s $28 default time frame, and the traded instrument is index futures
billion equity derivatives market and handles 75 percent of and not stock, we also made changed in the base line of ADX
the stock trades. (8) and changed it to 16. Below 16 we considered not to trade
and above 40, the trade is considered fully matured and
4.3 Introduction to Swing Trading caution is exercised.
Swing Trading takes advantage of brief price swings in 4.5.3 Strength in Signal
strongly trending stocks to ride the momentum in the direction
of the trend. [37] Generation of signal only after the trend is properly set by
1 Swing trading combines the best of two worlds -- the buying or shorting by bigger player in market. For e.g. FII,
slower pace of investing and the increased potential gains DII, FDI, LIC, Insurance companies, Mutual funds etc.
of day trading.

1059
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

Any trend is followed by a consolidation phase which consists

of profit booking and/or shorting activities. When a signal is
generated it should be ensured that it is backed by the
positions of bigger institutes like mentioned above. When
there is a break point in the price and market comes out of
consolidation phase, that breakout should be volume based,
else there is a high probability that the breakout will result in a
whipsaw. So until markets direction is set by position taken by
bigger players, our system should not generate a buy or sell
signal. We took proper note of this and EMAs in our system
are selected accordingly.

4.5.4 Using the Signals to their Full Capacity

As we backtested our system in the range of April 2005 to

march 2012, where we came across all the moods of nifty, we Figure 11. Chart for Exponential moving average
observed that there were 57% of successful trades and 43% of
loss making trades. As this ratio is close so it was necessary to SMA: 10 period sum / 10
use a trailing stoploss instead of fixed price target or to step by Multiplier: (2 / (Time periods + 1) ) = (2 / (10 + 1) ) = 0.1818
step booking of profit by playing in multiple numbers. After (18.18%)
different research methods we came across the idea to put EMA: {Close - EMA (previous day)} x multiplier + EMA
stoploss of a trailing EMA for all trades. Benefit of a trailing (previous day).
stoploss is that, it triggers by itself and there is no need to pre-
assume the price targets which involve human emotional A 10-period exponential moving average applies an 18.18%
interference. There are times when a trade goes long enough weighting to the most recent price. A 10-period EMA can also
beyond the imagination of anyone. It can only be ridded by not be called an 18.18% EMA. A 20-period EMA applies a 9.52%
fixing any target profit. Least the decision making of weighing to the most recent price (2/(20+1) = .0952). Notice
individual better will be the system. It’s a market saying that that the weighting for the shorter time period is more than the
80% of profits come from only 20% of trades. weighting for the longer time period. In fact, the weighting
drops by half every time the moving average period doubles.
The first parameter we selected is the most popular and
commonly used in high grade trading strategies, EMA. [32]
The Exponential Moving Average (EMA) weighs current
4.5.5 Removing Low Probability Trades prices more heavily than past prices. This gives the
Exponential Moving Average the advantage of being quicker
to respond to price fluctuations than a Simple Moving
Like MACD we have also used crossover of EMAs. But not
getting into complex calculation. We only took 2 EMAs, one Average; however, they are more prone to whipsaws too.
This was the main reason to select EMA before MA.
for shorter period and second for bigger trend. We tried to use
SAR (stop and reverse) strategy in market but found that
Formula:
buying in negative trend and shorting in positive trend does
EMA = Price (t) * k + EMA (y) * (1 – k)
not give satisfactory result in long run. So we decided to trade
Where t = today, y = yesterday, N = number of days in EMA,
only long trades by combinations of bigger EMA, Macd and
k = 2/(N+1)
ADX and to short only when trend is negative. This boosted
4.6.2 EMA Combination
our results.
Two different EMA’s were taken as a combination. They are
4.6 Explanation of Technical Parameters Used
the shorter -EMA (8), and the longer EMA (34).
Reason for choosing 8 and 34 ema is that 8 ema covers approx
4.6.1 EMA (EXPONENTIAL MOVING AVERAGE)
1.5 weeks trading days and 34 ema covers 1.5 month. So it’s a
combination of weekly and monthly emas. They are specially
A type of moving average that is similar to a simple moving optimized for trading the swing for index futures. These two
average, except that more weight is given to the latest data. parameters are widely used by day-traders and acts as strong
The exponential moving average is also known as support and resistance to price movements in index.
"exponentially weighted moving average". Here's a chart with The strategy includes the exemption of trades which lies
both an SMA and an EMA on it. between these two ema lines. The shorter ema reacts to faster
movements in index while the bigger decides the upward or
downward bias of index. All trades going against the trend are
skipped. Emas are combined because they give a slow entry

1060
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

and hence useful in avoiding the volatile times and also it -DI = current +DM = current positive
waits for a trend to ripe. negative directional
directional movement value
4.6.3 ADX (Average directional index) index
Ht = current high Ht-1 = previous high
An indicator used in technical analysis as an objective value
for the strength of trend. ADX is non-directional so it will Lt = current low Lt-1 = previous low
quantify a trend's strength regardless of whether it is up or - = current -DM = current negative
down. ADX is usually plotted in a chart window along with DMn modified directional
two lines known as the DMI (Directional Movement moving movement value
Indicators). ADX is derived from the relationship of the DMI average of -
lines. Analysis of ADX is a method of evaluating trend and DM
can help traders to choose the strongest trends and also how to TRn = current TRANSAC = true range
let profits run when the trend is strong. modified TION
moving
average of the
true range
N = number of DX = current DX
periods

4.6.4 MACD (Moving average convergence and divergence)

A trend-following momentum indicator that shows the

relationship between two moving averages of prices. The
MACD is calculated by subtracting the 26-day exponential
moving average (EMA) from the 12-day EMA. A nine-
day EMA of the MACD, called the "signal line", is then
plotted on top of the MACD, functioning as a trigger for buy
and sell signals.
The default settings for the MACD indicator are (26,12 and 9)
Ema:
• Slow moving average - 26 days
Figure 12. ADX • Fast moving average - 12 days
• MACD = 12 Day exponential moving average - 26 Day
Reason for using beside other indicator is that there are times exponential moving average
in market they are directionless. The combination of Emas
• Signal line - 9 day moving average of the difference
used above fails in those conditions and put a series of
between fast and slow.
whipsaws (false signals) which cannot be traded through them.
• All moving averages are exponential.
For this, we needed an indicator which gives the information
of direction of market. ADX does this well and the use of it
kept the strategy back on track by keeping out the trader from
entering into market.
We optimized ADX from its default parameter of 14 to 8 as
per the need of the system.
ADX = modify moving average of DX
DX = 100 x [ (+DI – (-DI) )/( +DI + (-DI)
)]
and
+DI = +DMn / TRn , -DI = -DM / TRn
+DM = Ht - Ht-1 , -DM = Lt - Lt-1
CL = Ct - Ct-1
TR = largest of +DM,-DM ,and CL.
Where

+DI = current +DMn = current

positive modified moving
directional average of +DM
index Figure 13. MACD

1061
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

4.7 Combining Trading Signals Using Neural Network 3. Positions are taken in last 5 min of market closing, if any
An Artificial Neural Network, often just called a signal/stoploss is generated.
neural network, is a mathematical model inspired by biological 4. All the stoplosses are considered as triggered only on the
neural networks. A closing basis.
5. No positions are taken in between, if a trade is missed.

Algorithm:
Begin
Buy Signal :
If index close is above the bigger EMA, and
ADX is not showing consolidation phase, and
Macd is Positive.
Stoploss :

This buy trade will be kept open till the index close is above
the shorter EMA, which acts as a trailing stop loss.

Triggering the stoploss, if index rises above the shorter EMA,

again a buy is created.
Figure 14. Neural Network Shorting Signal :
Neural network consists of an interconnected group If index close is below bigger EMA, and
ADX not showing consolidation, and
of artificial neurons, and it processes information using
a connectionist approach to computation. In most cases a Macd is Negative.
neural network is an adaptive system that changes its structure Stoploss :
during a learning phase. Neural networks are used to model This trade will be kept open till index closes above
shorter EMA, which acts as a trailing stop loss
complex relationships between inputs and outputs or to find
patterns in data. Triggering the stoploss, if index return below the shorter
EMA, again a short is created.
A trading strategy can consists of one or more
signals. Individual results of technical parameters don’t Every positions stoploss is the shorter EMA.
If ADX value in any occurrence is showing a directionless
outperform a stock, and they have a unique way to give some
indication about the stock. E.g. RSI (Relative strength index) value, the buy/short trade will be skipped.
If at any time, Macd or Adx is not supporting the signal, they
gives a overbought indication when it’s above 70, and
oversold when below 30. But this doesn’t mean that we won’t will be waited for the attunement and then only a signal will
be generated.
need any other indicator and only RSI is sufficient. Individual
results in backtesting RSI gave bad results, but it a well known End.
and popular indicator. So combining different signals and V. RESULTS AND OBSERVATIONS
trying to overcome one’s weakness by other, led to the concept
of automated trading strategy system.

Figure 15 . A snapshot of our trading system with signals and

technical parameters.

4.8 Procedure and Algorithm of Proposed Methodology Explanation of our System’s snapshot: The given chart is
subdivided into 3 parts. The 1st part shows the Nifty movement
1. Data of Index Future are collected from NSE Server 15 with add-ons of 8 and 34 Emas. It also shows where the
min before stock market closing and are taken as close of signals are generated and where their stoploss triggered.
that day. Signals are shown by arrows, green and red, and stoplosses are
2. Data is updated into the trading signal generating shown by green and red stars. The 2nd segment contains the
software.( E.g. Amibroker, Ninja Trader etc.) Adx signal and 3rd shows the Macd indicator.

1062
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

Observation: This chart explains about the number of positive

Index Point Returns Chart - Absolute and negative trade occurring in any strategy. A common
observation is that, in most strategies, the negative trades are
more than positive trades, which shows that all these
strategies, works on the principle of taking small losses to
generate big profits. They chose the higher probability trades
through technical analysis.

Se File Pro Tot Nega Pos % % Retu

ri Nam fit al tive itiv Re Re rn
al e in Tr Trad e tur tur on
Figure 16. Index Point Returns Chart - Absolute Poi ade es Tra n n 3000
nts s des per 0
Observation: The following chart shows that our
yea Inve
Proposed Methodology has outperformed all signals and
r stme
indicators including Nifty. Nifty futures have a lot size of
nt
50 shares, which can be taken as absolute return of any
1 Inter 63. 9.0 4890
system trading on index futures. Absolute return of Nifty
is 3228 points which is handsomely outperformed by our est 00 0% 0
Rate %
Proposed Methodology of 5410 points. ADX and
s
Bollinger bands individually outperform Nifty with
generating 3894 and 4151 points. 2 Mac 224 12 76 51 62. 8.8 1123
d 6 7 12 7% 00
%
3 Posit 279 46 22 24 65. 9.4 1397
ional 5 98 3% 50
Ema %
Cros
sover
4 Nifty 322 84 14 20. 1614
8 6.2 90 00
8% %
5 Adx 389 14 97 45 12 17. 1947
4 2 1.2 32 00
Figure 17. Return on Investment 5% %
6 Bolli 415 52 30 22 11 16. 2075
Observation: Trading with Proposed Methodology nger 1 3.6 23 50
generates a handsome returns of Rs 270500/- in 7 years on Band 2% %
investment of Rs 30000/-. A maximum brokerage of Rs s
15000/- can be considered. Short term capital tax 7 Shor 470 25 150 109 11 16. 2351
implications are not considered. All technical parameters ter 3 9 5.9 57 50
outperform FD (fixed deposit) interest rates with quite a Rang 6% %
margin. e
Ema
Cros
sover
8 Prop 541 14 83 62 13 19. 2705
osed 0 5 8.1 73 00
Meth 0% %
odol
ogy

Figure 19. Strategies and their performance

Observation and clarification:

This table gives all details of the schemes compared.
Figure 18. No. of Positive and Negative Trades The “% Return” column gives the absolute return from the
schemes but they can’t be compared because their bases are
different. For example, the starting position of nifty index was

1063
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

near 2000 and at end it reached 5000, so for calculations, 2000 [4] L. Lin, D. Luo, and L. Liu. Mining domain-driven
was taken as base for a single trade. As far as our Proposed correlations in stock markets. In S. Zhang and R. Jarvis,
Methodology is concerned, there are 145 trades of different editors, AI 2005: Advances in Artificial Intelligence:
18th Australian Joint Conference on Artificial
bases. So “Profit in Points” column gives a better idea of the
Intelligence, pages 979–982,2005
performance, where our methodology generated 5410 points [5] E. Fink and K. B. Pratt. Indexing of compressed time
as compared to 3228 generated by Nifty. So we conclude that series. InM. Last, A. Kandel, and H. Bunke, editors,
our Proposed Methodology outperforms Base Index - Nifty. Data Mining in Time Series Databases, volume 57 of
Machine Perception and Artificial Intelligence, pages
VI. CONCLUSION & FUTURE SCOPE 43–65. World Scientific, 2004[6] M. Kantardzic. Data
Mining – Concepts, Models, Methods, and Algorithms.
The simulation result and observations in chapter 4 can be IEEE Press, Piscataway, NJ, USA, 2003.
summarized as follows: Note that these observations are [6] M. Kantardzic. Data Mining – Concepts, Models,
restricted to the scope of the project which is using Index Methods, and Algorithms. IEEE Press, Piscataway, NJ,
Futures as trading instrument during the period of April 2005 USA, 2003.
to March 2012. [7] http://www.investopedia.com/active-trading/technical-
indicators/#axzz25hOUjQ00
1. From all technical signals, Bollinger bands signal exhibited [8] John L. Person “ A complete guide to technical trading
the highest annual return on investments. tactics “
2. Adx and Bollinger bands both outperformed the Nifty [9] Richard Dennis and William Eckhardt , “ The original
index but Macd could not. Proposed Methodology gave Turtle Trading Rules ”
highest returns on the absolute index point return basis. [10] http://en.wikipedia.org
3. Number of negative trades mostly exceeds the positives [11] http://www.ino.com, “17 Money making candlestick
trades in all strategies. patterns”
4. Automated trading shows better performance in long term [12] http://thebestbusinessintheworld.blogspot.in/search?q=
as compared to hit and trail methods. 34+ema
5. Volume of trading has increased significantly when SEBI [13] Nikhil Bakshi ETH Zurich Aug 2008, “ Stock market
permitted automated trading in India in 2005. prediction using online data : fundamental and
6. Combination of indicators results in better performance technical approaches”
than individual indicators which are unidirectional in their [14] Suhasini Subramanian Batch 2011, MSc Student of
approach. Economics and Finance Durham Business School,
Ustinov College Durham University United Kingdom
Future Scope: This Proposed Methodology does not include (UK), “The Impact of the Introduction of Index
the hedging strategies, and also it is not tested in “Nifty Futures on Volatility and Noise Trading ”
Options” which can reduce the loss generated in more than [15] Http://www.informedtrades.com/f400/
half of total trades. Future work could involve both these left [16] Darie Moldovan , Babes-Bolyal university of Cluj-
aspects. The third main addition can be the inclusion of Napoca Dec 2011 ,“Business intelligence – data mining
candlestick patterns. All these ideas are left for later versions in finance”
of this Proposed Methodology. [17] Http://www.swing-trade-stocks.com/swing-trading-
strategy.html
REFERENCE [18] Richi Nayak and Paul te Braak, Faculty of Information
technology, Queensland university of technology,
[1] M. C. Chan, H. C. Leung, and W. D. Luo, “ Brisbane, “ Temporal Pattern matching for the
Association mining system for financial ratios and prediction of stock prices ”
stock prices in China and Hong Kong stock exchange” [19] Http://stockcharts.com
In A. Abraham, Y. Dote, T. Furuhashi, M. K¨oppen, A. [20] “Accuracy Driven Artificial Neural Networks in Stock
Ohuchi, and Y. Ohsawa, editors, Soft Computing as market Prediction” Selvan Simon and Arun Raoot
Tran disciplinary Science and Technology, pages1151– Asst. Professor, P. G. Dept. of Computer Sci., SNDT
1160. Springer Verlag Berlin Heidelberg, 2005. Women’s University, Mumbai
[2] E. Fink and K. B. Pratt. Indexing of compressed time [email protected] Professor, National Institute
series. In M. Last, A. Kandel, and H. Bunke, editors, of Industrial Engineering, Vihar Lake, Mumbai.
Data Mining in Time Series Databases, volume 57 of [21] Monica Tirea Computer Science Department West
Machine Perception and Artificial Intelligence, pages University of Timisoara Timisoara, Romania
43–65. World Scientific, 2004. [email protected], “Multi-Agent Stock Trading
[3] L. Lin, L. Cao, J.Wang, and C. Zhang. The applications Algorithm Model”. Loan Tandau Green Mountain
of genetic algorithms in stock market data mining Analytics Cary, North Carolina USA
optimization. In Proceedings of Fifth International [email protected] and Viorel Negru Computer
Conference on Data Mining, Text Mining and their Science Department West University of Timisoara
Business Applications, pages 273–280, 2004 Timisoara, Romania [email protected]

1064
IJRITCC | August 2023, Available @ http://www.ijritcc.org
International Journal on Recent and Innovation Trends in Computing and Communication
ISSN: 2321-8169 Volume: 11 Issue: 9s
Article Received: 25 April 2023 Revised: 12 June 2023 Accepted: 30 July 2023
___________________________________________________________________________________________________________________

[22] K. Senthamarai Kannan, P. Sailapathi Sekar, M. [40] Fayyad et al., 1996. Piatetsky-Shapiro et al. 1991,
Mohamed Sathikand P. Arumugam, “Financial Stock Silberschatz et al.1995
Market Forecast using Data Mining Techniques” [41] Singh, A., Bhardwaj, G., Srivastava, A.P., Bindra, A.
[23] Savinderjit Kaur - Department of Information and Chaudhary, P., 2022, April. Application of neural
Technology, UIET,PU, Chandigarh, India and Veenu network to technical analysis of stock market
Mangat -Assistant Professor, department of prediction. In 2022 3rd International Conference on
Information Technology, UIET, PU, Chandigarh, India. Intelligent Engineering and Management (ICIEM) (pp.
“Applications of data mining in stock market “ 302-306). IEEE.
[24] The empirical study of applying technical analysis on [42] Liu, X., Guo, J., Wang, H. and Zhang, F., 2022.
DJI , HSI and Taiwan stock market Prediction of stock market index based on ISSA-BP
[25] Robert P. Schumaker and Hsinchun Chen Artificial neural network. Expert Systems with Applications, 204,
Intelligence Lab, Department of Management p.117604.
Information Systems The University of Arizona, [43] Napitupulu, H., Sambas, A., Murniati, A. and
Tucson, Arizona 85721, USA{rschumak, Kusumaningtyas, V.A., 2022. Artificial Neural
hchen}@eller.arizona.edu Network-Based Machine Learning Approach to Stock
[26] “ Stock Trend Prediction Using Regression Analysis – Market Prediction Model on the Indonesia Stock
A Data Mining Approach “ S Abdulsalam Sulaiman Exchange During the COVID-19. Engineering
Olaniyi, 2Adewole, Kayode S. , 3Jimoh, R. G 1Al- Letters, 30(3).
Hikmah University,Ilorin, Nigeria,2Kwara State [44] Sharma, D.K., Hota, H.S., Brown, K. and Handa, R.,
University, Malete, Nigeria,3University of Ilorin,Ilorin, 2022. Integration of genetic algorithm with artificial
[email protected], neural network for stock market
[email protected], forecasting. International Journal of System Assurance
[email protected] Engineering and Management, 13(Suppl 2), pp.828-
[27] Ayo Charles K., 1Adebiyi Marion O., and 2Otokiti 841.
Sunday O. Stock Price Prediction using Neural [45] Srivinay, Manujakshi, B.C., Kabadi, M.G. and Naik,
Network with Hybridized Market Indicators 1 Adebiyi N., 2022. A hybrid stock price prediction model based
Ayodele A.,1Department of Computer and Information on PRE and deep neural network. Data, 7(5), p.51.
sciences, Covenant University, Ota, 2 Department of [46] Manujakshi, B.C., Kabadi, M.G. and Naik, N., 2022. A
Business Studies, Covenant University, Ota, Nigeria hybrid stock price prediction model based on pre and
Author: VOL. 3, NO. 1, January 2012 ISSN 2079- deep neural network. Data, 7(5), p.51.
8407 Journal of Emerging Trends in Computing and [47] Sako, K., Mpinda, B.N. and Rodrigues, P.C., 2022.
Information Sciences. Neural networks for financial time series
[28] http://www.icharts.in/charts.html forecasting. Entropy, 24(5), p.657.
[29] Arun K. Pujari, “ Data Mining Techniques ” [48] Li, X., Wang, J., Tan, J., Ji, S. and Jia, H., 2022. A
[30] Nhamo Mdzingwa Supervisor :John Ebden “ Data graph neural network-based stock forecasting method
Mining with Oracle 10g using Clustering and utilizing multi-source heterogeneous data
Classification Algorithms : 30 May2005 : Rhodes fusion. Multimedia tools and applications, 81(30),
University COMPUTER SCIENCE HONOURS pp.43753-43775.
PROJECT [49] Li, C. and Qian, G., 2022. Stock Price Prediction Using
[31] Boris kovalerchuck central Washington University a Frequency Decomposition Based GRU Transformer
USA Evgenni Vityaev institute of mathematics, Neural Network. Applied Sciences, 13(1), p.222.
Russian academy of sciences, Russia : ”Data mining for
financial application”
[32] http://www.onlinetradingconcepts.com/TechnicalAnaly
sis/MAExponential.html
[33] Tutorial PPT by Richard J.Roiger, Michael Geatz. Data
Mining: A Tutorial Based Primer
[34] Discovery of Patterns and evaluation of Clustering
Algorithms in Social Network Data (Face book
100Universities) through Data Mining Techniques and
Methods Edelstein, 1999) Nan cy.P1, Dr.R. Geetha
Ramani2 1Ph.D Research
[35] Nitchi & Avram-Nitchi, 1997
[36] [Berry et al, 2000]
[37] [Pyle, 2000]
[38] http://www.tradersedgeindia.com/swing_trading.htm
[39] http://kalyan-city.blogspot.in/2010/11/what-is-stock-
exchange-its-definitions.html

1065
IJRITCC | August 2023, Available @ http://www.ijritcc.org