ADDIS ABABA SCIENCE AND TECHNOLOGY UNIVERSITY COLLEGE OF

ELECTRICAL & MECHANICAL ENGINEERING

Department of Software Engineering Big Data Analytics

Assignment
A Research Proposal On ATM Cash Demand Prediction Using Time
Series Approach:- A Case Study On Enat Bank

- Advanced Research Methods and Ethics -

Prepared by: Hiwot Teshome

PhD Student, AASTU

Abstract
One of the problems related to the banking system in Automated Teller Machine (ATM) is cash

demand forecasting. If an ATM faces a shortage of cash, it will face the decline of bank popularity

and in turn will have some costs and the bank will encounter decreasing customers' use of these

systems. When an ATM experiences a cash overflow, it is at risk of being charged for premium

insurance, resulting in unwanted payments. This can prevent the ATM from profiting from the cash

that is put into it.

So to avoid this Situation This research proposal focuses on forecasting cash demands at

automatic teller machines (ATMs) using Deep Learning Approach. The objective of this proposal

is to provide a predictive approach so that ATM operators can anticipate and plan for the cash

demands of their customers. This approach will help them to better manage their cash levels and

reduce the risk of running out of cash. The proposed research will analyze existing data to develop

a model that can accurately predict the cash demands of ATMs. The model will then be tested and

evaluated to ensure it is accurate and reliable. The results of this research will provide valuable

insights that can be used to improve the management of cash demands at ATMs.
Introduction
Due to the introduction of advanced technologies in the financial sector, banks are rendering their

services to customers using technologies. Among such technologies, the automated teller machine

(ATM) is the most popular one.

Automatic Teller Machines (ATMs) are devices used by financial institutions to support a group of

services round the clock in public space or area automatically. One of the major services these

ATMs offer is the disbursement of cash from a corresponding account of a customer.

In Ethiopia, the deployment of Automated Teller Machine (ATM) started in the late 2001.In the

2020/21 fiscal year, Ethiopia’s banking industry processed a total of 225,603,984 ATM

transactions, of which 38,302,370 were inter-bank transactions offered by EthSwitch S.C.

Banks would obtain benefits such as reserve funds, efficacy, more noteworthy buyer inclusion,

client satisfaction, and loyalty if they provide quality services through electronic managing account

networks such as ATM (Al-Hawari and Ward, 2006).

Enat Bank announced the start of its ATM deployment for the 2021/2022 fiscal year. This is a

major step forward for the bank, as it will enable customers to access their accounts and withdraw

cash from ATMs located in convenient locations. This deployment will make banking easier and

more convenient for customers, as they will no longer have to wait in line at the bank.

The deployment of ATMs was part of Enat Bank's commitment to providing customers with the

best possible service and experience. With this new deployment, customers can expect to have

access to their accounts and cash 24 hours a day, 7 days a week.Inorder to fulfill this requirements

ATM must be filled with the right amount so that neither a customer’s transaction is rejected

because of out-of-cash status, nor the idle cash ruins the opportunity for the bank to earn profit on

it.
In an ATM, while it would be devastating to run out of cash, it is important to keep cash at the

right levels to meet customer demand. In such cases, it becomes very necessary to have a

forecasting system in order to get a clear picture of demands in advance. Having the models to

forecast the daily (or weekly) cash demand for every ATM, it is possible to plan and to optimize

the cash loads for the whole ATM network.

Deep Learning modeling is a fairly popular area in the banking sector. One of these tasks is the

problem of predicting the optimal amount of loaded cash in ATMs with proper planning and cash

flow management. It is possible to minimize the costs of servicing cash flows, for example, by

increasing the interval between ATM collections (collection period), to reduce the costs of

collector services and to relieve the cash departments.

One of the main challenges in today’s banking industry is to forecast the cash demand of their

ATM network. Each ATM must be filled with the right amount so that neither a customer’s

transaction is rejected because of out-of-cash status, nor the idle cash ruins the opportunity for the

bank to earn profit on it. This paper proposed a time-series model for forecasting the cash demands

of each ATM in a network of ATMs for Enat Bank .

Deep Learning modeling and forecasting is a data analysis technique that involves building

empirical models using historical data to predict future values. This approach is chosen because

the sample data is time-consecutive and based on a specific time frame. By analyzing the observed

characteristics of the data, we can make informed predictions about future trends.
Research Objectives
The following objectives are the focus of the research:

General Objective

The General Objective of this research is to use Deep Learning Approach and Neural Network for

the forecasting model and Predict ATM Cash Demand.

Specific Objective

● To Compare existing Deep Learning Models for ATM cash demand forecasting using the

Actual Data Set

● Predict the ATM Cash Demand using the selected Model

● Evaluate the validity of the Prediction using the Validity Checker Mechanisms
Literature Review

There are many systematic reviews on specific applications of financial Deep Learning

forecasting. The dominance of forecasting financial and stock markets is traced in this area.

Several predictive stock market research reviews have been published based on various soft

computing methods at different times (Nair and Mohandas 2014; Cavalcante et al. 2016;

Atsalakis and Valavanis 2009). Chatterjee et al. (2000), Katarya and Mahajan (2017) focused

on ANN-based financial market predictive studies, while (Hu et al. 2015) focused on machine

learning (ML) implementations for inventory forecasting and algorithmic trading models. In

another application for Deep Learning forecasting, researchers have reviewed studies on

forex forecasting using ANN (Hewamalage et al. 2021) and various other soft computing

methods (Pradeepkumar and Ravi 2018)

ATM cash flow prediction is a challenging problem, and various approaches in literature give

varying degree of per- formance for this task, a handful of Neural Networks and a few

observing a Deep Learning approach and solving it with similar models like ARIMA, SARIMA,

and VARMAX. The root problem lies in understanding the trends, seasonality and exogenous

variables in a Deep Learning (Shcherbitsky et al. 2019)

Note that in all the variety of financial and economic applications of Deep Learning fore- casting

methods, insufficient attention is paid to more specific problems (Sezer et al. 2020), such as

forecasting the ATM load, which is the subject of this article. However, this does not mean that

this problem has not been considered in the literature (Ekinci et al. 2019). We note, for example,

the works of M. Rafi, which are devoted to modeling ATM loading using classical statistical
methods such as moving average autoregression (Rafi et al. 2020) and the use of recurrent neural

networks based on LSTM cells (Asad et al. 2020). LSTM can generally be considered a classical

method since it was created specifically for modeling time series. Therefore, it often finds

application in this problem (Arabani and Komleh 2019; Serengil and Ozpinar 2019; Rong and

Wang 2021). Also, classical and deep learning methods were applied to this problem, the

performance comparison of which are given in Vangala and Vadlamani (2020) and Hasheminejad

and Reisjafari (2017). However, an analysis of the literature shows that classical machine learning

methods such as regression analysis (Rajwani et al. 2017), support vector machines (Jadwal et al.

2018), dynamic pro- gramming (Ozer et al. 2019), ARIMA (Khanarsa and Sinapiromsaran 2017),

and gradient boosting (Shcherbitsky et al. 2019) are used much more often.

Mention should be made of the research work carried out based on the database of the NN5

competition (Adeodato et al. 2014), since this competition is very closely related to our work. The

dataset in this competition contains two years of daily cash withdrawals from 111 ATMs in the

UK. The challenge is to predict ATM withdrawals for each ATM over the next 56 days.

The most efficient model, from Andrawis et al. (2011), an ensemble of general Gaussian

regression, neural network, and linear models, achieved a SMAPE of 18.95%. The article by

Venkatesh et al. (2014) improved the prediction accuracy of Andrawis et al. (2011) by clustering

similar ATMs and applying many popular neural networks. These studies show that modern ML

algorithms for forecasting Deep Learning produce results better than classical

Many studies are devoted to applying machine learning methods for problems of forecasting

time series. The most developed areas of application include stock market forecasting (Aliev
et al. 2004), trading system development (Dymowa 2011), and practical examples of market

forecasting (Kovalerchuk and Vityaev 2006) using machine learning models. Bahrammirzaee

(2010) reviewed research on financial forecasting and planning and other financial

applications using various artificial intelligence techniques such as artificial neural networks

(ANNs), expert systems, and hybrid models.

For example, evolutionary algorithms find many applications in forecasting eco- nomic and

financial Deep Learning (Chen 2002); multipurpose evolutionary algorithms have been

widely studied for various economic applications, including Deep Learning forecasting (Tapia

and Coello Coello 2007; Ponsich et al. 2013; Aguilar-Rivera et al. 2015). Li and Ma examined

implementations for stock price forecasting and some other financial applications (Li and Ma

2010). Tkácˇ and Verner (2016) investigated various implementations of ANN in financial

applications, including for predicting stock prices. Recently Elmsili and Outtaj investigated

applications of ANNs in economic and management research, including series forecasting

(Elmsili and Outtaj 2018).

In the literature, techniques used for cash demand forecasting can be broadly classified into four

groups [5-8]: (1) Deep Learning method that predicts future cash need based on the past values of

variable and/or past errors. Generally, techniques in Deep Learning approach work well unless

there is an abrupt change in the environment or variables that are believed to affect the cash

withdrawal pattern. (2) Factor analysis method, which is based on the determination of various

factors that influence the cash demand pattern and calculating their correlation with actual cash

withdrawal. The purpose is to determine the functional form of this influence (independent

variables) and to use this to forecast future values of the dependent variables. However, estimating

this functional form is more difficult when they have a non-linear relationship. (3) Fuzzy expert
system approach that tries to imitate the reasoning of a human operator. The idea is to reduce the

analogical thinking behind the intuitive

Research Gap Literatures suggest that most of the studies have been done on issues related to

electronic banking (like internet banking, Mobile banking and ATM service). However, no work

has been done in Ethiopia with regard to ATM cash demand prediction issues. The present study

intends to assess ATM Cash Demand Enat Bank.

Research Methodology

Data collection and modeling are the two key parts of the methodology that this research study

uses.Real data is essential for our forecasting project to train our models to handle real-life

situations effectively. We will be using a quantitative research methodology and data from 50

ATMs, selecting 6 based on their transactional experience and network availability. Our analysis

will consider a 2-year period of cash withdrawals which will be used as training data for our model

and we will conduct an examination before preprocessing the datasets. We'll also employ

supervised learning methods like neural networks to build a model that can comprehend the

prediction of the cash on an ATM.

Expected Outcomes

At the conclusion of this research project, we anticipate achieving two primary results.

1) Predicted Cash Demand of the sample ATMs

2) Evaluated Deep Learning Model based on the Data Set

11