




































Indian Journal of Finance and Banking 

 Vol. 5, No. 2; 2021 

                                       ISSN 2574-6081   E-ISSN 2574-609X 

Published by CRIBFB, USA 

 

1 

NEURAL NETWORKS IN FINANCE: A DESCRIPTIVE 

SYSTEMATIC REVIEW  

 
 

Dr. K. Riyazahmed 

Assistant Professor 

Shri Dharmasthala Manjunatheshwara 

Institute for Management Development  

Mysore, Karnataka, India 

E-mail: riyazahmed@sdmimd.ac.in 

 

 

ABSTRACT 

Traditional statistical methods pose challenges in data analysis due to irregularity in the 

financial data. To improve accuracy, financial researchers use machine learning architectures 

for the past two decades. Neural Networks (NN) are a widely used architecture in financial 

research. Despite the wider usage, NN application in finance is yet to be well defined. Hence, 

this descriptive study classifies and examines the NN application in finance into four broad 

categories i.e., investment prediction, credit evaluation, financial distress, and other financial 

applications. Likewise, the review classifies the NN methods used under each category into 

standard, optimized and hybrid NN. Further, accuracy measures used by the research work 

widely differ, in turn, pose challenges for comparison of a NN under each category and reduces 

the scope of formalizing a theory to choose optimum network model under each category.  

 

Keywords: Neural Networks, ANN, Analytics, Machine Learning. 

 

JEL Classification Codes: G1, G17, M150. 

 

INTRODUCTION 

Financial data are immensely available, yet the innate nature of big data shows uncertainty, 

incompleteness, and inconsistency which pose challenges in using traditional statistical methods 

for financial data analysis (Brooks et al., 2019; Hariri et al., 2019). Financial researchers try to 

overcome the traditional statistical limitations by using machine learning architecture like Neural 

Networks (NN).  

NN imitates the human brain by using nodes and layers of connections, which pass 

signals with a set of associated weights and bias adjustments (Figure 1). NN results are not easily 

interpretable and so the analysis is in black-box nature. Irrespective of the non-interpretability of 

results, NN received importance in financial research due to its computing efficiency in handling 

financial big data. 

For the last two decades, financial researchers are using NN in various analyses like risk 

classification (Altman, Marco, & Varetto, 1994), bankruptcy and share price prediction (Barr & 

Mani, 1994). Despite the attention, NN application in finance is yet to be well defined. The last 

decade saw an increase in the finance research using NN in an analysis like forecasting of share 

prices (Chang, Liu, Fan, Lin, & Lai, 2009; Sapna & Argente, 2003; Safer, 2002), option prices 

mailto:E-mail:%20riyazahmed@sdmimd.ac.in


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

2 

(Lin & Yeh, 2009; Kohler, Krzyzak, & Todorovic, 2010), and the future prices (Dunis, Laws, & 

Evans, 2008; Laws & Dunis, 2013).  

 

 

 

 

 

 

 

 

 

 

 

 

Figure 1. standard Neural Network 

 

The Literature reviews of Feldman and Kingdon (1995), Wong and Selvi (1998), Vellido, 

Lisboa, and Vaughan (1999), Krishnaswamy, Gilbert, and Pashley (2000), Coakley and Brown 

(2000), Fadlalla and Lin (2001), Wei, Nakamori, Wang, and Yu (2007), Cavalcante et al. (2016), 

and Huang, Chai, and Cho (2020) are the existing works in this regard. Yet, the review works 

show significant limitations. Firstly, no review follows a protocol-based review process which is 

essential for reproducibility. Secondly, several reviews are not examining financial applications 

entirely or they focus on aspects like soft computing, and computational intelligence, instead of 

NN architectures (Table 1). Even though the studies analyze the NN applications to a certain 

extent, non-reproducibility is a serious concern.  

Further, the absence of a systematic review method results in serious drawbacks in the 

quality of review findings (Karunananthan, Maxwell, & Welch, 2020). Systematic reviews have 

greater potential than other research designs leading to the reproducibility of research 

(Shokraneh & Adams, 2019). Since the computing efficiency doubles every two years 

(Gustafson, 2011) which improves the efficiency of handling complex data sets, exploring the 

research works to date with the scientific review methods will help to understand the existing 

status of NN in analyzing the financial data.  

 

Table 1. Summary of existing reviews 

 

Study Period of 

review 

Nature of 

study 

Focus of the 

study 

Summary of conclusion 

Feldman and 

Kingdon (1995) 

1988-1996 

(*Authors‟ 

estimation) 

Descriptive, 

Non-

Systematic 

review. 

Advantages of 

MLP, BPNN, 

and SOM. 

Generalization, 

architecture selection, 

and application of 

selected NN. 

Wong and Selvi 

(1998) 

1990 – 1996 Descriptive, 

Review 

process 

disclosed.  

Classification 

of NN 

application in 

finance. 

The implication to NN 

developers. 

Velido, Lisboa, 1992 – 1998 Descriptive, Application of Comprehensively 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

3 

and Vaughan 

(1999) 

Review 

process 

disclosed. 

NN in 

business. 

reported the most quoted 

advantages and 

disadvantages of NN in 

various business 

applications. 

Krishnaswamy et 

al.  (2000) 

1989 – 1996 

(*Authors‟ 

estimation) 

Descriptive, 

Non – 

systematic 

review 

Description of 

NN and its 

finance 

application. 

Backpropagation NN has 

proven robust. 

Supervised and 

unsupervised NN is used 

in finance. 

Coakley and 

Brown (2000) 

1988 – 1997 

(*Authors‟ 

estimation) 

Descriptive, 

Non – 

systematic 

review 

Financial 

Application, 

development 

of ANN 

models. 

ANN researchers face a 

challenge that there are 

no formal theories for 

determining optimal 

network model 

Fadlalla and Lin 

(2001) 

1986 – 1997 Descriptive, 

Non – 

systematic 

review 

Financial 

Application, 

focus on 

feedforward 

and feed 

backward NN 

models. 

NN has great promise 

for financial applications 

and combinations of two 

approaches should be 

investigated. 

Calderon and 

Cheh (2002) 

1993 – 1999 Descriptive, 

sources of 

review 

disclosed. 

NN in 

auditing and 

risk 

management. 

NN shows promising 

performance in 

Preliminary analytical 

procedures in the 

auditing process. 

Wei Huang et al. 

(2007) 

NA Descriptive, 

Non – 

systematic 

review 

Focus on input 

variables, NN 

models 

applied in 

forex, stock 

market, and 

economic 

forecasting. 

The prediction 

performance of neural 

networks can be 

improved by being 

integrated with other 

technologies. 

Cavalcante et al. 

(2016) 

2009 -2015 Descriptive, 

review process 

disclosed 

Computational 

intelligence in 

finance (NN is 

a part of the 

study) 

Categorized studies into 

preprocessing, 

forecasting, and text 

mining. 

Huang et al. 

(2020) 

2014 – 2018 Descriptive, 

review 

collection 

process 

disclosed. 

Deep learning 

applications in 

finance and 

banking. 

Reports about data 

inputs, preprocessing 

and evaluation rules of 

deep learning in finance 

and banking 

 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

4 

Hence, this structured review reveals an interpretable pattern of NN architectures and their 

application in finance research. The study follows the systematic review process provided by 

Moher, Liberati, Tetzlaff, Altman, and The PRISMA Group (2009) and Gupta, Chauhan, and 

Jaiswal (2019). The primary aim of the study is to classify the research papers based on their NN 

application in finance research. This is done by, 

 Classification of major topics and sub-topics, and 

 Identification of various NN architectures used in the classified subtopics.  

 

METHOD 

A literature review starts with searching for quality research papers from prominent journals 

(Ngai & Wat, 2002). Further collecting research papers from the online database has become an 

emerging culture in the information era (Petter & Lean 2009). So, this study uses the EBSCO 

Business Elite database, which is a repository of 525 peer-reviewed research journals, to collect 

research papers. The systematic review process prescribed by Moher et al. (2009) comprises 

defining a protocol for literature search, exclusion criteria of research papers, and final selection 

of papers (Figure 2). 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2. Selection process of research papers (Moher et al., 2009) 

 

Protocol directs the research paper selection based on the criteria. To obtain research 

papers the study followed an advanced search option in the EBSCO Business Elite database and 

used two keywords „Neural Networks‟ and „Finance‟. Research studies published in English 

under the subject areas of business, management, and finance are only considered. Empirical 

articles that are published in peer-reviewed academic journals are collected at the first level.  

Restricting the review only to published articles can strengthen quality control since 

many of the academic journals follow meticulous publishing criteria in terms of research 

contribution and robustness of the results (Light & Pillemer, 1984). The protocol process has 

helped to collect papers with high research quality. After collection, the studies with incomplete 

details, irrelevant context, and duplicate publications are excluded.  

Excluding research papers for incomplete details, 

duplicate publications, irrelevant context, native language 

Total N = 51 

Scholarly Peer-reviewed articles 

[Excluding magazines, cover stories] 

N = 131 

Primary search: EBSCO Host Business Elite 

N = 141 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

5 

After exclusions, data extraction is done by carefully considering the title, abstract, and 

overall theme of the paper focuses on applying NN in finance. 141 papers are collected from the 

EBSCO Business Elite database and during the first level of screening 77 research papers are 

excluded. Finally, 51 research papers are considered for review.  

 

RESULTS 

The researcher scrutinized the collected research articles for their relevancy and suitability to be 

considered as a part of this review paper. When research papers fulfill the established criteria, the 

author read the full paper to find its contributions. Figure 3 represents the broader classification 

of research papers under major topics and sub-topics.   

 

 
IP – Investment Prediction; CE – Credit Evaluation; FD – Financial Distress. 

Figure 3. Classification of research papers based on major topics and sub-topics 

 

Based on the reading the author identified four main topics and several subtopics (Table 

2). The next section discusses each main topic, inferences of the research work carried in the 

subtopics based on the NN methods. Further, NN with statistical and architectural advancements 

is classified as 'Optimized NN'. NNs incorporating financial theories and knowledge are 

classified under 'Hybrid NN'. A model-free NN is classified as 'Standard NN'.  

 

Table 2.  Major research topics and subtopics 

 

Main topics Sub-topics 

Investment prediction Prediction of options prices, futures prices, share prices, Forex, 

indexes, bond yields, commodity spreads, trading patterns, 

shareholder wealth, and portfolio performance.  

Credit evaluation Predicting credit risk, and  

Estimating credit rating. 

Financial distress Evaluating financial distress 

Other financial applications Development of financial intelligent system,  

Detecting fraudulent financial reporting, 

Assessment of systematic risk,  

Evaluating operating performance,  

Assessing project portfolio performance.  



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

6 

 
Figure 4.  Framework of NN application in finance 

 

DISCUSSION 

 

Investment Prediction: 

Derivatives -  

a) Options  

Optimized NN –  

Back Propagation NN (BPNN) – BPNN minimizes the prediction error by giving the nodes with 

higher error rates lower weights and vice versa. Hence, BPNN is found to be more suitable for 

derivative prediction (Kaastra & Boyd, 1995). When used to predict Taiwan stock index options, 

BPNN demonstrated improved accuracy in support of hedging at in-the-money option (Lin & 

Yeh, 2009).  

Likewise, the research study of Hutchison et al. (1994) used BPNN to predict the prices 

of S & P 500 futures and options, yet the result is that BPNN did not show significantly better 

accuracy than other linear models like ordinary least squares.  

 

NN with Monte Carlo simulation – Monte Carlo, is a simulation technique to understand the 

impact of risk and uncertainty in prediction. It is used with linear NN for predicting American 

options (Kohler et al., 2010).  Since it is a simulation based NN, the accuracy would widely 

differ in empirical prediction.  

 

Advanced Modular NN (AMNN) – AMNN is a series of independent NN which serves as a 

module and operates on separate inputs to accomplish a subtask. AMNN gives more accuracy 

than the standard NN when predicting European call option prices (Gradojevic et al., 2009). 

 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

7 

Feed forward NN (FFNN) – FFNN is used when the nature of financial data is neither 

sequential nor time-dependent. FFNN predicts European Index options and S&P 500 European 

call options more accurately than standard NN (Gencay & Gibson, 2007).  

 

Hybrid NN -  
NN with Black Scholes - The research study by Blynski, and Faseruk (2006) compared the 

effectiveness of option price forecasting using the traditional Black Scholes model with NN 

(hybrid NN) and standard NN. Likewise, the research study by Chen and Sutcliffe (2012) 

confirms that a hybrid NN along with Black Scholes predicts accurately than the standard NN or 

Black Scholes model individually. Similarly, Sperckelsen et al. (2014) used Black Scholes model 

variables in predicting option prices of currency futures (EUR/USD and concluded that the 

hybrid model is better than the theoretical option pricing model and MLP.   

 

NN with Black Scholes & Wavelet – Zapart (2003) in his research study uses Wavelet along 

with NN and Black Scholes. Wavelet is a mathematical advancement that addresses oscillations 

that decays quickly in a data set. The study found that Black Scholes NN with Wavelet predicts 

superior when analyzing option prices.  

 

Hybrid Black Scholes NN with stochastic volatility - Stochastic volatility represents the nature 

of volatility fluctuating over time. The research study by Gencay and Gibson (2007) found that 

the hybrid NN model with stochastic volatility predicts better than the standard NN Model while 

predicting European stock index options.   

 

b) Futures –  

Standard NN – Model-free NN, without the attributes of financial theories, also performs 

significantly in the case of predicting currency futures prices. A model-free NN is used to predict 

the high-frequency currency futures and the predictability power is better than the closed-form 

financial model (Sperckelsen et al., 2014). 

 

Optimized NN -  
Multilayer perceptron (MLP) – MLP commonly represents a feed-forward NN with three 

layers. MLP is used to predict the commodity futures to hedge against corn and ethanol spreads 

and found to be accurate in prediction (Dunis et al., 2015). Likewise, Karathanasapoulos et al. 

(2016) used MLP in gasoline futures contracts.  

 

Higher-order NN (HONN) -HONN utilizes a higher combination of NN inputs. The research 

study of Dunis et al. (2015) compared the performance of HONN with MLP. The study 

concludes that MLP and HONN are superior in predicting with leveraging option. Sermpinis et 

al. (2013) tested HONN in predicting index futures. Likewise, Karathanasapoulos et al. (2016), 

in their research study used HONN to predict gasoline futures contracts. 

 

Radical base function neural network (RBFNN)– Karathanasapoulos et al. (2016) used a 

radical basic function neural network (RBFNN) to improve the trading performance of futures. 

RBFNN transforms the input signal into another form, which can be then feed into the network 

to get linear separability. The study concludes that RBF NN is superior in both trading 

performance and statistical accuracy. 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

8 

Hybrid NN -  

Feature transformed NN – Kim (2004) used feature transformed NN, including domain-specific 

factors like relative strength index to predict the futures prices and found that feature transformed 

NN predicts accurately than the linear models and concludes that incorporating domain 

knowledge in NN architecture improves performance.  

 

Shares & Indices –  

Standard NN –  

Haefke and Helmenstien (2002) used NN in forecasting indices and inferred that applicability of 

information criteria is important in the selection of NN. In contrast, Moreno and Olmeda (2007) 

claim that NN is not superior in predicting the stock markets to the linear models.  

 

Optimized NN -  

NN with Statistical optimization – LV D et al. (2020) used principal component analysis (PCA), 

Least absolute shrinkage and selection operators (LASSO), classification and regression trees 

(CART), and Piecewise linear representation methods (PLR) to optimize NN. As a result, there 

no significant improvement in NN incorporating the features, however, NN with PLR, resulted in 

an improvement in profit through better forecast ability. 

 

Back propagation (BPNN) and Piecewise linear representation (PLR) – In a research study by 

Chang et al. (2009), it is found that BPNN along with PLR consistently created good results for 

predicting upward, steady, and downward trends of stock prices.  

 

Hybrid NN –  

NN with Fama French five-factor model - Besides customizing the NN, researchers have used 

financial models like Fama and French five-factor model with NN and found improvement in the 

profitability of investors in both linear and nonlinear data (Jan & Ayub, 2019). 

 

NN with GARCH model– Ozbey and Paksoy (2020) combined GARCH with NN and compared 

the performance of the hybrid model with the classic GARCH model. The study found that the 

hybrid model is superior in predicting volatility to the classic GARCH.  

 

NN with Top-down theory, technical analysis, and dynamic time series methods –Huang, 

G.,Huang,GB., Shiji, and Youa (2014) used integrated models using conventional top-down 

trading theory, technical analysis, and dynamic time series methods and concludes that the 

hybrid system gives remarkable investment returns and demonstrates promising potential tools 

for stock market forecasting. 

 

Other related financial forecasts –  

Optimized NN –  

Back propagation NN (BPNN) –Chiang et al. (1996) used BPNN to predict the Net Asset Value 

(NAV) of mutual funds and found that BPNN outperforms the regression model. Jain and Nag 

(1995) predicted the prices of initial public offering (IPO) using BPNN and found significant 

economic benefits in BPNN. 

 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

9 

Multi-layer perception (MLP) – Indroa et al. (1999) used MLP and compared it with stepwise 

linear regression. The results show that MLP is superior to the linear model.  

 

General regression NN (GRNN) –Barnes and Lee (2009) used GRNN in analyzing the effect of 

macroeconomic and firm-specific factors in determining shareholder wealth.  

 

Multivariate NN -Wie et al. (2004) used Multivariate NN to predict earning per share in 

comparison with univariate and multivariate linear models incorporating fundamental accounting 

variables and found that NN along with accounting variables predicts more accurately than linear 

forecasting models.  

 

Table 3. NN research in Investment prediction  

 

 Author(s) Purpose NN Model Sample Period Predictors Comparison 

R Squared 

Value Accuracy 

1 

Kaastra and 

Boyd (1995) 

Forecastin

g 

Economic 

time-

series data 

Back 

Propagation 

NN 

Conceptu

al Paper - - - - - 

2 

Lin and Yeh 

(2009) 

Forecastin

g Option 

Prices 

Back 

Propagation 

NN 

15 582 

call 

option 

price data 

points 

2003 - 

2004 

Black 

Scholes 

variables  - 

MAPE 

5.2534 

3 

Hutchison et 

al. (1994) 

Pricing 

and 

Hedging 

derivative 

securities 

Ordinary 

Least squares, 

Radical Basis 

functions 

network, 

Multi-layer 

Perceptron, 

Projection 

Pursuit 

S&P 500 

Future 

and 

options 

1987 - 

1991 

Black 

Scholes 

variables 

No 

significant 

difference 

between 

models 84.76 - 

4 

Kohler et al. 

(2010) 

Pricing of 

American 

Options 

Least square 

NN  

Monte 

Carlo 

Simulated 

Data - - - - - 

5 

Gradojevic 

et al.  (2009) 

Pricing 

European 

Call 

options 

Modular NN 

& Black 

Scholes NN 

 S&P-500 

index 

European 

call 

option 

prices, 

Chicago 

Board 

Options 

Exchange 

1987 - 

1994 

Black 

Scholes 

variables 

BS NN 

Model > 

Modular NN - 

MAPE 

1.87 

6 

Gencay and 

Gibson 

(2007) 

Pricing 

European 

Stock 

Index 

options 

Feedforward 

NN 

S&P 500 

index9 

options 

from the 

Berkeley 

1989 - 

1991 

Price of the 

underlying, 

strike price, 

volatility, 

interest 

FFNN > 

Stochastic 

volatility 

(SV) and 

stochastic - - 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

10 

Option 

Database 

rate, time 

to maturity 

volatility 

random jump 

(SVJ),  

7 

Blensky and 

Fasurek 

(2006) 

Comparin

g Option 

prices 

forecast of 

NN with 

Black 

Scholes 

model  

Back 

Propagation 

NN 

64, 280 

OEX 100 

Index call 

option 

1986 – 

1993 - 

NN > Black 

Scholes 

model - - 

8 

Chen and 

Sutcliffe 

(2012) 

Pricing 

and 

hedging 

short 

sterling 

options 

Standard NN, 

Modified 

Black Model 

NN, and 

Hybrid NN 

Short 

sterling 

futures 

traded on 

NYSE 

2012 

(Quarterl

y expiry 

cycle) 

Ask, bid, 

trade, 

spread 

trade, and 

block trade, 

Hybrid NN > 

Modified 

Black model, 

Standard NN r = 98.6864 - 

9 

Sperckelsen 

et al. (2014) 

Realtime 

pricing of 

options on 

currency 

futures 

Model-free 

option pricing 

NN, Multi-

Layer 

Perceptron 

EUR/US

D option 

on 

currency 

future 

from 

Chicago 

Mercantil

e 

Exchange 

(CME) 2012 

Futures 

price, 

Strike 

price, 

Expiration 

time, Risk-

free rate, 

Asset 

volatility 

Hybrid NN > 

theoretical 

option pricing 

model 99% 

MAPE 

0.3146 

1

0 

Zapart 

(2003) 

Pricing 

European 

and 

American 

Call 

options 

NN with 

Binomial 

trees and 

Wavelets, NN 

with Genetic 

algorithm and 

Black Scholes 

model 

Options 

prices as 

quoted on 

the 

Chicago 

Board 

Options 

Exchange 

are used 2003 

Time to 

expiry, 

current 

stock price, 

risk-free 

rate - - - 

1

1 

Dunis et al. 

(2015) 

Modeling 

corn/ethan

ol crush 

spread 

MLP, HONN, 

GPA 

Ethanol 

futures 

contract 

traded in 

Chicago 

Board 

2005 - 

2010 Leverage 

GPA > 

HONN, MLP - - 

1

3 

Sermpinis et 

al.  (2013) 

The 

forecastin

g FTSE 

100 

futures 

Higher-order 

NN, Multi-

Layer 

Perceptron, 

Recurrent 

neural 

networks 

FTSE 100 

futures 

2007 - 

2008 

Realized 

daily 

returns (21 

days) 

HONN> 

MLP, RNN - 18.85% 

1

4 

Karathanaso

poulos et al. 

(2016) 

Modeling 

crack 

spread 

RBF, PSO, 

MLP - 

2005  - 

2015 

20 ARIMA 

and 10 

GARCH 

models 

PSORBF > 

MLP - - 

1 Kim (2004) Future Feature Korean May to Positive Feature - - 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

11 

5 price 

prediction 

transformed 

ANN based 

on domain 

knowledge  

stock 

index 

(KOSPI) 

Novemb

er 1996 

volume 

index, Rate 

of Change, 

Momentum

, etc. 

transformed 

ANN > 

Linear ANN 

1

6 

Haefke and 

Helmenstien 

(2002) 

Index 

Forecastin

g and 

Model 

Selection 

Feedforward 

NN 

Austrian 

Traded 

Index 

(ATX) 

2 

Novemb

er 

1992 to 

14 

October 

1994 

Geometric 

mean, 

arithmetic 

mean 

The proposed 

integrated 

model shows 

significant 

performanc 0.041 

AMAPE = 

1.862 

1

7 

Moreno and 

Olmeda 

(2007) 

Predictabi

lity of 

emerging 

and 

developed 

stock 

markets 

using NN Standard NN 

 49 MSCI 

(Morgan 

Stanley 

Capital 

Internatio

nal) 

indexes 

March 

1995 to 

March of 

2001 

(1560 

daily 

observati

ons) 

index 

returns, 

daily and 

weekly 

NN is not 

superior to 

the linear 

models - - 

1

8 

LV D et al. 

(2020) 

Dimensio

nality 

reduction 

in stock 

trading 

 MLP, Deep 

Belief 

Network 

(DBN), 

Stacked 

Auto-

Encoders 

(SAE), RNN, 

LSTM, Gated 

Recurrent 

Unit (GRU) 

 US 

SPICS 

and the 

Chinese 

CSICS 

Past 

2000 

trading 

days of 

SPICS 

and 

CSICS 

before 

Decembe

r 

31, 2017 

44 

technical, 

Volatility, 

Psychologi

cal, cash 

flow 

indicators. LASSO NN - - 

1

9 

Chang et al. 

(2009) 

Stock 

Trading 

Points 

Prediction 

BPNN, GA, 

PLR 

Stock 

Prices 

2004/01/

02 to 

2006/04/

12 

Moving 

average, 

Bias, RSI, 

ninety days 

stochastic 

line, etc. 

PLR +GA 

improves 

Profitability - - 

2

0 

Jan and 

Ayub (2019) 

Improving 

the 

predictabil

ity of 

Fama 

French 

five-factor 

model Standard NN 

Manufact

uring 

companie

s in 

Pakistan 

Stock 

Exchange 

2000 to 

2015 

Market 

cap, 

BV/MV 

ratio, % in 

total assets, 

and EBIT  

NN improves 

FF model r = 0.99989 

MSE = 

0.0012 

2

1 

Ozbey and 

Paksoy 

(2020) 

Estimatio

n of index 

returns 

with 

GARCH 

and NN 

Hybrid NN, 

Exp GARCH, 

and Nor. 

Distrn 

Borsa 

Istanbul 

100 price 

Index  

2017 - 

2018 

Borsa 

Istanbul 

100 Index 

value 

Hybrid NN + 

GARCH > 

Hybrid NN + 

normal 

distribution - 

MSE = 

0.015926 

2

2 

Huang, 

Huang, 

Shiji, and 

Youa (2014) 

Integrated 

data 

mining in 

stock 

Top-down 

trading theory 

+ ANN + 

technical 

Taiwan 

Semicond

uctor 

Manufact

2011 - 

2013 

Stochastic 

KD, 

William 

%R, RSI, 

Integrated 

NN model 

improves 

stock - 

True 

positive = 

98.50% 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

12 

forecastin

g 

analysis + 

dynamic time 

series + and 

Bayesian 

probability 

uring 

Company 

and 

Evergreen 

Marine 

Corporati

on 

PSY line, 

ADX, MA, 

MACD 

forecasting 

2

3 

Chiang et al.  

(1996) 

Mutual 

Fund 

NAV 

forecastin

g BPNN 

6 Year 

economic 

variables 

and 101 

US 

mutual 

funds 

1981 - 

1986 

GNP, 

Consumpti

on, 

Investment, 

CPI, 

Money 

supply, 

unemploy

ment, T-

bill, Long 

term rate 

BPNN > 

Linear & 

Non-Linear 

regression 0.989 

MAPE = 

8.76 

2

4 

Jain and Nag 

(1995) 

Predicting 

IPO 

pricing FFNN 

552 IPOs 

in the 

United 

States 

1980 - 

1990 

11 

variables 

[Size, 

Underwrite

r, sales, 

ROA, ROI, 

Assets 

turnover, 

etc.]  - - - 

2

5 

Indroa et al. 

(1999) 

Predicting 

mutual 

fund 

performan

ce MLP 

Morningst

ar Mutual 

Funds 

On-Disc 

database 

1993 - 

1995 

Annualized 

return, 

turnover, 

P/E, P/B, 

Mar.Cap 

MLP > 

Linear 

models - 

MAPE = 

4.88 

2

6 

Barnes and 

Lee (2009) 

Effects of 

Macroeco

nomic-

Firm-

Specific 

Factors 

on 

Sharehold

er Wealth 

General 

regression 

NN (GRNN) 

Miscellan

eous 

Industrials 

in 

the 

Australian 

Stock 

Market 2007 

D/E, Gross 

margin, 

Debt to 

cash, EVA, 

EPS, 

WACC 

funds, 

ROIC  

ANN is 

effective in 

the prediction 0.0548 

MAE = 

37.649 

2

7 

Wie et al. 

(2004) 

NN model 

for EPS 

forecastin

g 

Univariate-

NN and 

multivariate 

NN 

Quarterly 

EPS of 

283 

companie

s in SEC 

1992 – 

2002 

Inventory, 

A/R, 

capital 

expenditure

, gross 

margin, 

Sel.Adm 

exp, Tax 

rate, labour 

force 

NN models > 

Linear 

models - 

MAPE = 

0.362 

 

 

 

 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

13 

Credit Risk Prediction: 

Credit evaluation –  

Optimized NN –  

Bayesian regularized NN (BRNN) - Sariev and Germano (2020) used BRNN to Predict the 

Probability of Default and found BRNN superior in prediction. 

 

Back propagation NN (BPNN) – Loss-given default (LGD) is used in credit risk assessment. 

LGD is the share of an asset that is lost if a borrower default. Loterman et al. (2012) compared 

the nonlinear techniques with the linear counterparts in predicting LGD of major international 

banks. 

 

Feed forward NN (FFNN) – Qi and Zhao (2011) found that nonparametric method like FFNN 

and regression tree predicts the LGD accurately both in and out of sample than their parametric 

counterparts. Cifter et al. (2009) investigated the relationship between industrial production and 

credit defaults (nonperforming loans) using FFNN based on wavelet decomposition. 

 

Fuzzy mathematical model –Aiqun et al. (2020) applied NN in risk assessment of logistic 

finance using back propagation NN and Fuzzy mathematical model. The study found that NN 

with the fuzzy mathematical model is accurate in risk assessment. Further, Baesens et al. (2003) 

provided a table with a graphical format that facilitates easy consultation to interpret the NN 

results. 

 

Credit Scoring – 

Standard NN -  

Chikolwa and Chan (2008) compared NN with ordinal regression (OR) to study the determinants 

of Commercial Mortgage-Backed Securities (CMBS) and concluded that NN is superior in 

prediction to OR. Trinkle and Baldwin (2007) applied NN in credit evaluation for loan finance 

and concluded that NN can be used in the credit scoring process with caution because of its 

hidden nature.  

 

Optimized NN –  

Backpropagation NN– Hajek (2011) applied NN to rate the United States municipalities in the 

state of Connecticut and found a higher accuracy of NN in classifying the municipalities with a 

limited subset of variables. Zan et al. (2004) compared the performance of Support vector 

machines (SVM) with back propagation NN (BPNN) on the credit rating of companies and 

found that both SVM and BPNN have the same accuracy in predicting the credit rating. 

 

Table 4. NN literatures in Credit Risk analysis 

 

 Author(s) Purpose NN Model Sample 

Output 

Variable Predictors Comparison 

R 

Squared 

Value Accuracy 

1 

Sariev and 

Germano 

(2020) 

Estimatio

n of the 

probabilit

y of 

default 

BRNN, 

BPNN 

East 

Europea

n, 

German, 

and 

Polish 

2007 - 

2012 

(EE), 

2007 - 

2013 (P) 

Payables 

turnover, 

ROA, cash 

ratio, 

sales/total 

assets, 

BRNN > 

BPNN - - 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

14 

data LA/TA, 

interest 

coverage 

2 

Lotterman

et al. 

(2012)# 

Benchmar

king 

regression 

algorithms 

for loss 

given 

default 

modeling 

NN, SVM, 

and OLS. 

six LGD 

datasets 

from 

internati

onal 

banks - - 

SVM, NN > 

Linear Models 0.1295 

MAE = 

0.3118 

3 

Qi and 

Zhao 

(2011)# 

Comparis

on of 

modeling 

methods 

for Loss 

Given 

Default 

OLS, 

fractional 

response 

regression 

(FRR), 

inverse 

Gaussian 

regression 

(IGR), and 

inverse 

Gaussian 

regression 

with beta 

transforma

tion (IGR-

BT), and 

regression 

tree (RT), 

NN 

3751 

defaulte

d 

securitie

s in the 

US, 

Moody‟

s 

Ultimate 

Recover

y 

Databas

e 

 1985 to 

2008 - 

NN, 

regression tree 

> linear 

regression, 

fractional 

response, OLS 0.576 - 

4 

Cifter et 

al. (2009) 

Examine 

the 

relationshi

p between 

industrial 

productio

n and 

credit 

defaults 

FFNN 

using 

Wavelet 

decomposi

tion 

83 

monthly 

observat

ions 

Industria

l 

producti

on and 

credit 

default 

rates are 

from 

Central 

Bank of 

Turkey 

2001 to 

2007 

Industrial 

production, 

Credit 

defaults all 

sectors, 

Wholesale, 

and retail 

trade 

Industrial 

cycle affects 

the sectoral 

credit-default 

cycles at 

different. 

time scales - 

MSE = 

0.00022 

5 

Aigun et 

al. (2020) 

Risk 

assessmen

t of 

logistic 

finance 

BPNN and 

Fuzzy 

mathemati

cal  - 2019 - 

NN +Fuzzy is 

accurate in the 

prediction - - 

6 

Baesens et 

al. (2003) 

Rule 

Extraction 

and 

Decision 

Tables 

 MLP, 

Neuro 

rule, 

Trepan, 

and Nef 

German 

credit 

dataset 

from 

UCI - 

Term of 

loan, 

Purpose, 

savings 

account 

Extract very 

compact rule 

sets and trees 

for all data 

sets - - 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

15 

for Credit-

Risk 

Evaluatio

n 

class. repositor

y, Bene1 

and 

Bene2 

datasets 

from 

Benelue

x 

financial 

institutio

ns 

balance, 

income, 

property, 

No. of years 

as a client, 

Economical 

sector 

7 

Chikolwa 

and Chan 

(2008) 

Determina

nts of 

credit 

ratings 

Standard 

NN Vs 

Ordinal 

regression 

 MBS 

credit 

ratings 

of 

Standard 

and Poor 

1999 - 

2005 

Loan to 

value; Debt 

Service 

Coverage 

Ratio; issue 

size; bond 

tenure, 

property 

diversity, 

geographica

l diversity, 

CMBS 

rating 

ANN > 

Ordinal 

regression 

Pseudo R 

squared = 

0.018 

Classificati

on 

accuracy = 

80% 

8 

Trinkle 

and 

Baldwin 

(2007) 

Interpreta

ble credit 

model 

developm

ent using 

NN 

ANN 

models 

created 

from 

previous 

research 

studies 

Two 

German 

consume

r credit 

data 

sets, 

SAS 

data 

repositor

y - 

Age, car, 

cards, Cash, 

etc. 

ANN > 

General credit 

models - 

Accuracy 

rate = 

0.63084 

9 

Hajek 

(2011) 

Municipal 

credit 

rating 

modeling 

by neural 

networks 

FFNN, 

RBFNN, 

Probabilist

ic NN, 

Cascade 

correlation 

NN, Group 

method of 

data 

handling 

(GMDH) 

polynomia

l NNs, 

Support 

Vector 

Machines. 

Credit 

informat

ion‟s of 

r 169 

US 

municip

alities 

(located 

in the 

State of 

Connect

icut) 

2003 -

2007 

Population. 

Population 

growth, 

median 

family 

income, 

unemploym

ent rate, 

total 

revenue to 

total 

expenditure, 

tax revenue 

to total 

revenue, tax 

collectibles, 

debt 

service, 

total debt to 

the total 

population, 

tax 

An accurate 

credit rating 

classification - 

PNN, 

Classificati

on 

Accuracy 

test = 

98.8% 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

16 

collection 

rate, form 

of income. 

1

0 

Zan et al. 

(2004) 

Credit 

rating 

analysis 

with 

support 

vector 

machines 

and neural 

networks 

backpropa

gation 

neural 

network 

(BNN) Vs 

Support 

vector 

machines 

Taiwan 

Ratings 

Corporat

ion,  

Securitie

s and 

Futures 

Institute, 

S&P 

Compus

tat data 

set. 

1991 to 

2000 

TA, TL, 

DE, CR, 

ROA, ROE, 

EPS, NOI, 

NII, etc. 

BNN, SVM 

>Linear 

regression - 

Accuracy 

rate = 80% 

# Cross validation 

 

Financial Distress: 

Optimized NN –  

Multi-layer perceptron (MLP) –Loukeris and Eleftheriadis (2015) used MLP, hybrid MLP with 

neurogenetic, and voted perceptron algorithm (VPA). VPA is a method that linearly separates 

data with a larger margin to predict financial distress. Manel (2012) used five MLP and 

compared them with traditional financial analysis to predict financial distress and found that the 

MLP is superior in accuracy.  

 

Learning vector quantization (LVQ) – LVQ is a NN method with a supervised algorithm to let 

choose the number of training instances to hang on to. Brockett et al. (2006) compared the 

multiple discriminant analysis (MDA) and logistic regression with LVQ to analyze the solvency 

of life insurance companies and found NN architectures are superior in predictions. 

 

Hidden layer learning vector quantization – Like LVQ, hidden layer LVQ (HDLVQ) 

outperforms the traditional NN methods and financial techniques while evaluating the corporate 

solvency of life insurance companies. The research study by Neves and Vieira (2006) integrated 

HDLVQ to correct the outputs of MLP and found the technique is superior to traditional 

techniques like z core models and standard NN. 

 
Hybrid NN –  

Fuzzy analytical hierarchy and CAMEL model – CAMELS framework is the most widely 

applied methodology to study the financial position of banks. Wanke et al. (2016) used NN with 

a fuzzy analytical hierarchical model along the CAMELS framework to predict the financial 

distress of banks. 

 

Z score model –Z score model is a financial technique to evaluate the financial distress of a 

company. Pradhan (2011) used NN along with Z score and found it classifies accurately. 

 

Profitability index and capital structure variables –Willer et al. (2020) created a business 

insolvency forecasting model using NN and found that the predictable power of NN shows 

significant accuracy. Yang et al. (1998) and Atiya (2001) confirms the accuracy.  



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

17 

Table 5.  NN literatures in Financial distress 

 
 

 

 Author(s) Purpose 

NN 

Model Sample 

Output 

Variable Predictors Comparison 

R 

Squared 

Value Accuracy 

1 

Loukeris 

and 

Eleftheria

dis (2015) 

Credit 

portfolio 

selection 

process 

MLP, 

hybrid 

MLP 

with 

neurog

enetic, 

and 

voted 

percept

ron 

algorith

m 

1411 

compan

ies from 

Greek 

commer

cial 

bank 

1994 to 

1997 

16 Financial 

ratios - - MSE 0.034 

2 

Manel 

(2012) 

Predictio

n of 

financial 

distress BPNN 

528 

Tunisia

n firm 

(Central 

bank of 

Tunisia 

report)   1999 - 2006 

26 financial 

ratios predicting 

distress - - 

Classificati

on 

accuracy – 

98.9% 

3 

Brockett 

et al. 

(2006) 

Compari

son of 

NN and 

statistical 

models 

for life 

insurers' 

financial 

distress 

predictio

n 

back-

propag

ation 

and 

learnin

g 

vector 

quantiz

ation 

(LVQ) 

Vs 

multipl

e 

discrim

inant 

analysi

s and 

logistic 

regressi

on 

analysi

s 

Texas 

Depart

ment of 

Insuran

ce data 

1991 to 

1995 IRIS variables 

BNN, LVQ > 

MDA, 

Logistic 

Regression - 

Correct rate 

(1994) = 

LVQ 

(1.00), BP 

(0.971) 

4 

Neves and 

Vieira 

(2006)# 

Improvin

g 

Bankrupt

cy 

Predictio

n 

with 

Hidden 

Layer 

Learnin

g 

vector 

quantiz

ation + 

Multi-

layer 

percept

ron 

780,000 

financia

l 

stateme

nts of 

French 

compan

ies, 

Industri 1998 - 2000 

Input consists of 

30 financial 

ratios   

Generalisat

ion error: 

MLP = 

8.8%, 

HLVC-Q = 

7.3% 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

18 

Learning 

Vector 

Quantiza

tion 

al 

French 

firms, 

583 

bankrup

t firms 

5 

Wanke et 

al. (2016) 

Predictin

g 

performa

nce in 

ASEAN 

banks 

Fuzzy 

analytic 

hierarc

hy 

process 

NN 

Financi

al ratios 

of 88 

Associa

tion of 

Southea

st Asian 

Nations 

bank 

2010 to 

2013 CAMELS ratios 

Possible to 

explain the 

causes of 

inefficiency 

using NN - 

RMSE = 

0.0248 

6 

Pradhan 

(2011) 

Predictio

n of 

financial 

distress BPNN 

State 

Bank of 

India 2001 – 2010 Z score variables - - - 

7 

Willer et 

al. (2020) 

Forecasti

ng 

business 

insolvenc

y 

Standar

d NN - - 

Profitability 

index, capital 

structure  - - 

91% 

Accuracy 

Dynamic 

Model 

8 

Yang et 

al. (1998) 

Probabili

stic 

Neural 

Network

s in 

Bankrupt

cy 

Predictio

n 

Fisher 

discrim

inant 

analysi

s, 

BPNN, 

PNN, 

PNN 

without 

patterns 

normali

zed. 

122 

compan

ies U.S. 

oil and 

gas 

industry 

1984 to 

1989 

Net cash flow to 

total assets, 

TD/TA, CL/TD, 

etc 

Fischer 

discriminant 

analysis, 

PNN 

normalized>   

BP NN, 

Probabilistic 

NN without a 

pattern  - 

Fisher 

discriminan

t analysis = 

87% 

correct 

classificati

on, 

9 

Atiya 

(2001) 

Bankrupt

cy 

Predictio

n for 

Credit 

Risk  

Standar

d NN 

Defaulte

d and 

from 

solvent 

US 

firms, 

716 

solvent 

firms 

and 195 

defaulted 

firm - 

Merton‟s asset-

based model  - - 

Correct rate 

= 85.50% 

# Cross validation 

 

Other Financial Applications: 

Standard NN 

Apart from the major topics of research, researchers have found NN is efficient to detect 

fraudulent reporting (Koskivaara & Back, 2007; Omar et al. 2017) and project portfolio 

management (Costantino et al., 2015). 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

19 

Optimized NN –  

A fuzzy analytical model is used for creating a financial information system (Wang et al., 2020). 

 

A self-organization map algorithm, an unsupervised learning NN that produces a low-

dimensional, discretized representation of the input space of the training samples, is called a 

map. SOM is used to analyze the integration of EU capital markets (Horobet 2014), differences 

in world economies (Cimpoeru 2015).   

 

Hybrid NN –  

NN with the Ohlson model is used to predict operating performance (Ying-Hua & Shih-Chin, 

2013).  

 

CONCLUSION 

A descriptive systematic review was conducted to find the application of neural networks in 

financial research. The study found a keen research interest to use NN for predicting financial 

data. This is obvious from the statistic that about 53% of the collected research studies applied 

NN in investment prediction. Credit evaluation and financial distress topics contribute to 20% 

and 17% of each of the collected papers. There are very few works (10%) found on other 

financial aspects.  

The following are the reflections of the review. First, it is observed that the researchers 

have used arbitrary data partition and architecture selection in all the research works. Besides, 

the performance or evaluation metrics widely differ among the collected research studies, giving 

less scope for comparing the accuracy of an NN architecture in a particular subtopic. Hence, in 

this study, a meta-analytic comparison to generalize the NN architectures and formalizing a 

theory to choose a suitable NN method under a topic has serious limitations. 

Second, there are few studies (Neves & Vieira, 2006; Qi & Zhao, 2011) that have 

performed the cross-validation in NN models. Cross-validation increases efficiency in using 

financial data as every observation is used for both training and testing which results in a more 

accurate estimate of out-of-sample prediction. Further, overfitting and underfitting of data will be 

managed efficiently through cross-validation.  

Third, unlike prediction, in the research studies of classifying financial data, there is a 

scope for a meta-analysis based on generalizing Area Under Curve (AUC) that help to estimate 

the accuracy of classification on a particular topic. Further, the study observed NN architectures 

including domain-specific knowledge performs with more accuracy. Hence, more domain-based 

Hybrid NN architectures can be trained.   

Besides, the review has the following limitations. The descriptive systematic review has 

examined only the research papers published and available under the EBSCO database. 

Consequently, the works of the literature review are prone to publication bias, which occurs with 

publishing only statistically significant results. Beyond, the research works in conference 

proceedings and working papers are not reviewed. Hence some sub-topics and main research 

topics might have been remaining uncovered. Systematically including more research papers 

from other sources will improve the chance of a meta-analysis of NN in financial research. Since 

meta-analysis on machine learning by Krittanawong et al. (2020) and Roelofs et al. (2019) are 

the only works available, that too in the medical domain, a meta-analysis of NN architectures in 

financial research will be a significant contribution to the existing financial and NN literature.   

 



https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

20 

REFERENCES 

Altman, E., Marco, G., & Varetto, F. (1994). Corporate Distress Diagnosis: Comparisons Using       

Linear Discriminant Analysis and Neural Networks (the Italian Experience). Journal of 

Banking and Finance, 18(3), 505-529. 

 

Aiqun, W., Zicong, H., Yilin, W., Kolivand, H., Balas, V. E., Paul, A., & Ramachandran, V. 

(2020). Risk assessment of logistics finance enterprises based on BP neural network and 

fuzzy mathematical model. Journal of Intelligent & Fuzzy Systems, 39(4), 5915–5925. 

https://doi.org/10.3233/JIFS-189066 

 

Atiya, A.F. (2001). Bankruptcy prediction for credit risk using neural networks: A survey and 

new results, IEEE Transactions on Neural Networks, 12(1), 929-935. 

 

Baesens, B., Setiono, R., Mues, C., & Vanthienen, J. (2003). Using Neural Network Rule 

Extraction and Decision Tables for Credit-Risk Evaluation. Management Science, 49(3), 

312–329. https://doi.org/10.1287/mnsc.49.3.312.12739 

 

Barr, D., & Mani. G. (1994). Using neural nets to manage investments, AI EXPERT, 34(3), 16–

22. 

 

Barnes, B, M., & Lee, C, S, V. (2009). Effects of Macroeconomic and Firm-Specific Factors on 

Shareholder Wealth: Some Australian Evidence. The Journal of wealth management, 12 

(1), 41-61. https://doi.org/10.3905/jwm.2009.12.1.041 

 

Brockett, P. L., Golden, L. L., Jang, J., & Yang, C. (2006). A Comparison of Neural Network, 

Statistical Methods, and Variable Choice for Life Insurers‟ Financial Distress Prediction. 

Journal of Risk & Insurance, 73(3), 397–419. https://doi.org/10.1111/j.1539-

6975.2006.00181.x 
 

Brooks, C., Andreas, G, F., Hoepner, McMillan, D., Vivian, A, & Simen, C, W. (2019) 

Financial data science: the birth of a new financial research paradigm complementing 

econometrics? The European Journal of Finance, 25(17), 1627-1636, 

https://doi.org/10.1080/1351847X.2019.1662822 

 

Blynski, L., & Faseruk, A. (2006). Comparison of the effectiveness of option price forecasting: 

Black-Scholes vs. simple and hybrid neural networks. Journal of Financial Management 

& Analysis, 19(2), 46. 
 
Cavalcantea, R, C., Brailerio, C, R., Souza, L, V., Nobrega, P, J., Oliveria, L, A. (2016). 

Computational Intelligence and financial markets: A survey and future directions. Expert 

systems with applications, 55, 194 – 211. http://dx.doi.org/10.1016/j.eswa.2016.02.006 

 

Chang, P. C., Liu, C. H., Fan, C. Y., Lin, J. L., & Lai, C. M. (2009). An Ensemble of Neural 

Networks for Stock Trading Decision Making. In: Huang DS., Jo KH., Lee HH., Kang 

HJ., Bevilacqua V. (eds) Emerging Intelligent Computing Technology and Applications. 

With Aspects of Artificial Intelligence. ICIC 2009. Lecture Notes in Computer Science, 

5755. https://doi.org/10.1007/978-3-642-04020-7_1 

https://doi.org/10.3233/JIFS-189066
https://doi.org/10.1287/mnsc.49.3.312.12739
https://doi.org/10.3905/jwm.2009.12.1.041
https://doi.org/10.1111/j.1539-6975.2006.00181.x
https://doi.org/10.1111/j.1539-6975.2006.00181.x
https://doi.org/10.1080/1351847X.2019.1662822
http://dx.doi.org/10.1016/j.eswa.2016.02.006
https://doi.org/10.1007/978-3-642-04020-7_1


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

21 

Chen, F., & Sutcliffe, C. (2012). Pricing and Hedging Short Sterling Options Using Neural 

Networks. Intelligent Systems in Accounting, Finance & Management, 19(2), 128–149. 

https://doi.org/10.1002/isaf.336 
 

Chikolwa, B., & Chan, F. (2008). Determinants of Commercial Mortgage-Backed Securities 

Credit Ratings: Australian Evidence. International Journal of Strategic Property 

Management, 12(2), 69–94. https://doi.org/10.3846/1648-715X.2008.12.69-94 

 

Chiang, W.C., Urban, T. L. & Baldridge, G.W. (1996). A neural network approach to mutual 

fund net asset value forecasting, Omega - International journal of management science, 

24, 205–215. 

 

Cifter, A., Yilmazer, S., & Cifter, E. (2009). Analysis of sectoral credit default cycle 

dependency with wavelet networks: Evidence from Turkey. Economic Modelling, 26(6), 

1382–1388. https://doi.org/10.1016/j.econmod.2009.07.014 

 

Cimpoeru, S. (2017). Using Self organizing maps for assessing systemic risk. Evidences form 

global economic crisis. The Bucharest University of Economic Studies. 

http://www.ecocyb.ase.ro/nr20152/05%20-%20Cimpoeru_Smaranda.pdf 

 

Coakley, J., & Brown. C. (2000). Artificial neural networks in accounting and finance: 

Modelling issues. Intelligent Systems in Accounting Finance & Management,9(2). 

 

Costnatino, F., Gravio, G, D., & Nonino, F. (2015). Project selection in project portfolio 

management: An artificial neural network model based on critical success factors. 

International journal of project management, 33, 1744 – 1754. 

http://dx.doi.org/10.1016/j.ijproman.2015.07.003  

 

Dunis, C. L., Laws, J., Middleton, P. W., & Karathanasopoulos, A. (2015). Trading and hedging 

the corn/ethanol crush spread using time-varying leverage and nonlinear models. 

European Journal of Finance, 21(4), 352–375. 

https://doi.org/10.1080/1351847X.2013.830140 

 

Fadlalla, A., & Lin, C. (2001) An analysis of the applications of neural networks in finance. 

Interfaces, 31(4), 112–122. 

 

Feldman, K, & Kingdon, J. (1995). Neural networks and some applications to finance, Applied 

Mathematical Finance, 2(1), 17-42. https://doi.org/10.1080/13504869500000002 
 

Gençay, R., & Gibson, R. (2007). Model Risk for European-Style Stock Index Options. IEEE 

Transactions on Neural Networks, 18(1), 193–202. 

https://doi.org/10.1109/TNN.2006.883005 

 

Gradojevic, N., Gençay, R., & Kukolj, D. (2009). Option Pricing With Modular Neural 

Networks. IEEE Transactions on Neural Networks, 20(4), 626–637. 

https://doi.org/10.1109/TNN.2008.2011130 

https://doi.org/10.1002/isaf.336
https://doi.org/10.3846/1648-715X.2008.12.69-94
https://doi.org/10.1016/j.econmod.2009.07.014
http://www.ecocyb.ase.ro/nr20152/05%20-%20Cimpoeru_Smaranda.pdf
https://doi.org/10.1080/1351847X.2013.830140
https://doi.org/10.1080/13504869500000002
https://doi.org/10.1109/TNN.2006.883005
https://doi.org/10.1109/TNN.2008.2011130


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

22 

Gupta, P., Chauhan, S. & Jaiswal, M.P. (2019). Classification of Smart City Research - a 

Descriptive Literature Review and Future Research Agenda. Information System 

Frontiers, 21, 661–685. https://doi.org/10.1007/s10796-019-09911-3 

 

Gustafson, J., L. (2011). Moore‟s Law. In: Padua D. Encyclopedia of Parallel Computing. 

Springer, Boston, MA. https://doi.org/10.1007/978-0-387-09766-4_81 

 

Hutchison A.W, Lo A.W, & Poggio, T. (1994) A nonparametric approach to pricing and 

hedging derivative securities via learning networks. The journal of Finance, 49(3), 851-

889. 

 

Hajek, P. (2011). Municipal credit rating modelling by neural networks. Decision support 

systems, 51 (1), 108 – 118. https://doi.org/10.1016/j.dss.2010.11.033 

 

Hariri, R. H., Fredericks, E.M. & Bowers, K. M. (2019). Uncertainty in big data analytics: 

survey, opportunities, and challenges. Journal of Big Data,6. 

https://doi.org/10.1186/s40537-019-0206-3. 

 

Haefke, C., & Helmenstein, C. (2002). Index forecasting and model selection. International 

Journal of Intelligent Systems in Accounting Finance & Management, 11(2), 119–135. 

https://doi.org/10.1002/isaf.214 

 

Horobet, A., Belascu, L., Ionita, I., & Serban - Oprescu, A.-T. (2014). A Neural Networks 

Perspective on the Financial integration of European Capital Markets. Economic 

Computation & Economic Cybernetics Studies & Research, 48(1), 1–12 

 

Huang, G., Huang, G, B., Shiji, S., & Youa, K. (2014). Trends in extreme learning machines: A 

review. Neural Networks. 61, 34 – 45. http://dx.doi.org/10.1016/j.neunet.2014.10.001 

 

Huang, J., Chai, J. & Cho, S. (2020). Deep learning in finance and banking: A literature review 

and classification. Frontiersof Business Research in China, 14. 

https://doi.org/10.1186/s11782-020-00082-6 

 

Indro, D. C., Jiang, C. X., Patuwo, B. E., & Zhang, G. P. (1999). Predicting mutual fund 

performance using artificial neural networks. Omega, 27(3), 373-380. 

https://doi.org/10.1016/S0305-0483(98)00048-6 
 

Jan, M. N., & Ayub, U. (2019). Do the Fama and French Five-Factor Model Forecast Well 

Using Ann? Journal of Business Economics & Management, 20(1), 168–191. 

https://doi.org/10.3846/jbem.2019.8250 

 

Kaastra, I. & Boyd, M. (1996). Designing a Neural Network for Forecasting Financial and 

Economic Time Series. Neurocomputing, 10, 215-236. https://doi.org/10.1016/0925-

2312(95)00039-9 

 

https://doi.org/10.1007/s10796-019-09911-3
https://doi.org/10.1007/978-0-387-09766-4_81
https://doi.org/10.1016/j.dss.2010.11.033
https://doi.org/10.1186/s40537-019-0206-3
https://doi.org/10.1002/isaf.214
http://dx.doi.org/10.1016/j.neunet.2014.10.001
https://doi.org/10.1186/s11782-020-00082-6
https://doi.org/10.1016/S0305-0483(98)00048-6
https://doi.org/10.3846/jbem.2019.8250
https://doi.org/10.1016/0925-2312(95)00039-9
https://doi.org/10.1016/0925-2312(95)00039-9


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

23 

Karathanasopoulos, A., Dunis, C., & Khalil, S. (2016). Modeling, forecasting, and trading with 

a new sliding window approach: the crack spread example. Quantitative Finance, 

16(12), 1875–1886. https://doi.org/10.1080/14697688.2016.1211796 

 

Karunananthan, S., Maxwell, L, J., & Welch, V. (2020) Protocol: When and how to replicate 

systematic reviews. Campbell Systematic Reviews. 16. https://doi.org/10.1002/cl2.1087 

 

Krishnaswamy, C.R., Gilbert, E.W., & Pashley, M.M. (2000) Neural network applications in 

finance: A practical introduction. Financial Practice and Education, 10(1), 75-84. 

 

Kriteenawong, C., Virk, H, H., Bangalore, S., Wang, Z., Johnson, W, K., Rachel. (2020). 

Machine learning prediction in cardiovascular diseases: a meta-analysis. Scientific 

reports, 10. https://doi.org/10.1038/s41598-020-72685-1 

 

Kohler, M., Krzyżak, A., & Todorovic, N. (2010). Pricing of High-Dimensional American 

Options by Neural Networks. Mathematical Finance, 20(3), 383–410. 

https://doi.org/10.1111/j.1467-9965.2010.00404.x 

 

Koskivaara, E., & Back, B. (2007). Artificial Neural Network Assistant (ANNA) for 

Continuous Auditing and Monitoring of Financial Data. Journal of Emerging 

Technologies in Accounting, 4, 29–45. https://doi.org/10.2308/jeta.2007.4.1.29 

 

Kim, K. (2004). Artificial neural networks with feature transformation based on domain 

knowledge for the prediction of stock index futures. Intelligent Systems in Accounting, 

Finance & Management, 12(3), 167–176. https://doi.org/10.1002/isaf.252 

 

Light, R.J., & Pillemer, D. B. (1984). Summing up: The science of reviewing research. 

Cambridge, MA: Harvard University Press. 

 

Lin, C.-T., & Yeh, H.-Y. (2009). Empirical of the Taiwan stock index option price forecasting 

model - applied artificial neural network. Applied Economics, 41(15), 1965–1972. 

https://doi.org/10.1080/00036840601131672 

 

Loterman, G., Brown, I., Martens, D., Mues, C., & Baesens, B. (2012). Benchmarking 

regression algorithms for loss given default modeling. International Journal of 

Forecasting, 28(1), 161–170. https://doi.org/10.1016/j.ijforecast.2011.01.006 

 

Loukeris, N., & Eleftheriadis, I. (2015). Further Higher Moments in Portfolio Selection and A 

Priori Detection of Bankruptcy, Under Multi-layer Perceptron Neural Networks, Hybrid 

Neuro-genetic MLPs, and the Voted Perceptron. International Journal of Finance & 

Economics, 20(4), 341–361. https://doi.org/10.1002/ijfe.1521 

 
LV, D., Wang, D., Li, M., & Xiang, Y. (2020). DNN models based on dimensionality reduction 

for stock trading. Intelligent data analysis, 24(1), 19-45.  https://doi.org/10.3233/IDA-

184403 

 

https://doi.org/10.1080/14697688.2016.1211796
https://doi.org/10.1002/cl2.1087
https://doi.org/10.1038/s41598-020-72685-1
https://doi.org/10.1111/j.1467-9965.2010.00404.x
https://doi.org/10.2308/jeta.2007.4.1.29
https://doi.org/10.1002/isaf.252
https://doi.org/10.1080/00036840601131672
https://doi.org/10.1016/j.ijforecast.2011.01.006
https://doi.org/10.1002/ijfe.1521
https://doi.org/10.3233/IDA-184403
https://doi.org/10.3233/IDA-184403


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

24 

Manel, H. (2012). Prediction of Financial Distress for Tunisian Firms: A Comparative Study 

between Financial Analysis and Neuronal Analysis. Business Intelligence Journal, 5(2), 

374–382. 

 

Moher, D., Liberati, A., Tetzlaff, J., Altman, D, G., & The PRISMA Group. (2009). Preferred 

Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. 

PLoS Med. 6(7). https://doi.org/10.1371/journal.pmed1000097 

 

Moreno, D., & Olmeda, I. (2007). Is the predictability of emerging and developed stock markets 

really exploitable? European Journal of Operational Research, 182(1), 436–454. 

https://doi.org/10.1016/j.ejor.2006.07.032 

 

Ngai, E. W. T., & Wat, F. K. T. (2002) A Literature Review and Classification of Electronic 

Commerce Research. Information & Management, 39, 415-429. 

 

Neves, J. C., & Vieira, A. (2006). Improving bankruptcy prediction with Hidden Layer Learning 

Vector Quantization. European Accounting Review, 15(2), 253–271. 

https://doi.org/10.1080/09638180600555016 

 

Omar, N., Johari, Z, A., & Smith, M. (2017). Predicting fraudulent financial reporting using 

artificial neural network. Journal of Financial Crime, 24(2), 362–387. 

https://doi.org/10.1108/JFC-11-2015-0061 

 

Ozbey, F., & Paksoy, S. (2020). Estimation of the XU100 Index Return Volatility with the 

Integration of GARCH Family Models and ANN. Business and economics research 

journal, 11(2), 385 – 396.   

 

Petter, S., McLean, E. R. (2009). A meta-analytic assessment of the DeLone and McLean IS 

success model: An examination of IS success at the individual level. Information 

Management, 46(3), 159-166. 

 

Pradhan, R., Pathak, K. K., & Singh, V. P. (2011). Z Score Reveals Credit Capacity: A Case 

Study of SBI. International Journal of Financial Management, 1(3), 72–78. 

 

Qi, M., & Zhao, X. (2011). Comparison of modeling methods for Loss Given Default. Journal 

of banking and finance, 35(11). 2842-2855. 

https://doi.org/10.1016/j.jbankfin.2011.03.011 

 

Roelofs, R., Shankar, V., Recht, B., Fridovich-Keil, S., Hardt, M., Miller, J & Schmidt, L. 

(2019). A Meta-Analysis of Overfitting in Machine Learning. Advances in Neural 

Information Processing Systems 32 (Neur IPS 2019), 1-11. Retrieved from 

https://openreview.net/pdf/98fdb4bb89fd868a5dda9f827133eec6e11bb4fa.pdf 

 

Safer, A. M. (2002). The application of neural networks to predict abnormal stock returns using 

insider trading data. Applied Stochastic Models in Business & Industry, 18(4), 381–389. 

https://doi.org/10.1002/asmb.466 

https://doi.org/10.1371/journal.pmed1000097
https://doi.org/10.1016/j.ejor.2006.07.032
https://doi.org/10.1080/09638180600555016
https://doi.org/10.1108/JFC-11-2015-0061
https://doi.org/10.1016/j.jbankfin.2011.03.011
https://openreview.net/pdf/98fdb4bb89fd868a5dda9f827133eec6e11bb4fa.pdf
https://doi.org/10.1002/asmb.466


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

25 

Sapna, O., Botti, V., & Argente, E. (2003). Application of neural networks to stock prediction in 

“pool” companies. Application of Artificial Intelligence 17(7):661–673. 

 

Sariev, E., & Germano, G. (2020) Bayesian regularized artificial neural networks for the 

estimation of the probability of default. Quantitative Finance, 20(2), 311–328. 

https://doi.org/10.1080/14697688.2019.1633014 

 

Sermpinis, G., Laws, J., & Dunis, C. L. (2013). Modeling and trading the realized volatility of 

the FTSE100 futures with higher-order neural networks. European Journal of Finance, 

19(3), 165–179. https://doi.org/10.1080/1351847X.2011.606990 

 

Shokraneh, F., & Adams, C. E. (2019). Study-based registers reduce waste in systematic 

reviewing: discussion and case report. Systematic Reviews, 8(1), 129. 

 

Sperckelsen, C., Mettenheim, H., & Breitner, M. H. (2014). Real-Time Pricing and Hedging of 

Options on Currency Futures with Artificial Neural Networks. Journal of Forecasting, 

33(6), 419–432. https://doi.org/10.1002/for.2311 

 

Trinkle, B. S., & Baldwin, A. A. (2007). Interpretable credit model development via artificial 

neural networks. Intelligent Systems in Accounting, Finance & Management, 15(3/4), 

123–147. https://doi.org/10.1002/isaf.289 

 

Vellido A., Lisboa, P. J., Vaughan, J. (1999). Neural networks in business: a survey of 

applications (1992–1998). Expert Syst Appl. 17(1), 51–70. 

 

Wie Zhang, A., Qing Cao, & Schniederjans, M. J. (2004). Neural Network Earnings Per Share 

Forecasting Models: A Comparative Analysis of Alternative Methods. Decision 

Sciences, 35(2), 205–237. https://doi.org/10.1111/j.00117315.2004.02674.x 

 

Wanke, P., Kalam Azad, M. A., Barros, C. P., & Hadi, V. A. (2016). Predicting performance in 

ASEAN banks: an integrated fuzzy MCDM-neural network approach. Expert Systems, 

33(3), 213–229. https://doi.org/10.1111/exsy.12144 

 

Wang, A., Liu, Y., Isaeva, E., & Rocha, Á. (2020). Intelligent financial management of 

company based on neural network and fuzzy volatility evaluation. Journal of Intelligent 

& Fuzzy Systems, 38(6), 7215–7228. https://doi.org/10.3233/JIFS-179798 

 

Wei, K., Nakamori, Y., Wang, S., & Yu, L. (2007). Neural Networks in Finance and Economics 

Forecasting. International Journal of Information Technology &Decision Making, 6(1), 

113-140. https://doi.org/10.1142/S021962200700237X 

 

Willer P.J., Giarola, V, A., Cardoso, C, A., & Pinheiro, N., T. (2020). Business Insolvency 

Forecasting Using Artificial Neural Networks. Management and Development, 17(2), 

136–162. https://doi.org/10.25112/rgd.v17i2.1777 

 

https://doi.org/10.1080/14697688.2019.1633014
https://doi.org/10.1080/1351847X.2011.606990
https://doi.org/10.1002/for.2311
https://doi.org/10.1002/isaf.289
https://doi.org/10.1111/j.00117315.2004.02674.x
https://doi.org/10.1111/exsy.12144
https://doi.org/10.3233/JIFS-179798
https://doi.org/10.1142/S021962200700237X
https://doi.org/10.25112/rgd.v17i2.1777


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

26 

Wong, B. K., & Selvi, Y. (1998). Neural network applications in finance: A review and analysis 

of literature (1990-1996). Information and Management, 34(3), 129–139. 

 

Yang, Z. R., Platt, M. B., Platt, H. D (1999). Probabilistic neural networks in bankruptcy 

prediction. Journal of Business Research, 44(2), 67–74. 

 

Ying-Hua, C., & Shih-Chin, W. (2013). Integration of Evolutionary Computing and Equity 

Valuation Models to Forecast Stock Values Based on Data Mining. Asia Pacific 

Management Review, 18(1), 63–78. https://doi.org/10.6126/APMR.2013.18.1.04 

 

Zan, H., Chen, H., Hsu, C., Chen, W., & Wu, S. (2004). Credit rating analysis with support 

vector machines and neural networks: a market comparative study. Decision Support 

Systems, 37(4), 543–558. https://doi.org/10.1016/s0167-9236(03)00086-1 

 

Zapart, C. A. (2003). Beyond Black–Scholes: A Neural Networks-Based Approach to Options 

Pricing. International Journal of Theoretical & Applied Finance, 6(5), 469. 

https://doi.org/10.1142/S0219024903002006 

 

 

APPENDICES 

 

Appendix A. Figure depicting NN architectures in Investment Prediction 

 

 
*Hybrid NN methods are developed by incorporating the financial theories (Black Scholes 

option pricing, Fama French five-factor model), advanced statistics (GARCH, PCA, LASSO, 

CART, PLR),  advanced NN architectures (HONN, GPA, MLP), and domain-specific factors. 

 

 

 

 

 

 

0 2 4 6 8 10 12 14 16

Standard NN

Back Propagation

Feed forward

Polynomial 

Monte carlo

Baysian regularisation

Hybrid NN

Wavelet

Radical Based function NN

Multi layer perceptron

Investment Prediction

https://doi.org/10.6126/APMR.2013.18.1.04
https://doi.org/10.1016/s0167-9236(03)00086-1
https://doi.org/10.1142/S0219024903002006


https://www.cribfb.com/journal/index.php/ijfb                           Indian Journal of Finance and Banking                         Vol. 5, No. 2; 2021 

27 

Appendix B. Figure depicting NN architectures in credit evaluation, financial distress, and other 

applications 

 

 
*Support Vector Machine is a similar machine learning algorithm like NN. 

 

 

Copyrights 

Copyright for this article is retained by the author(s), with first publication rights granted to the 

journal. This is an open-access article distributed under the terms and conditions of the Creative 

Commons Attribution license (http://creativecommons.org/licenses/by/4.0/) 
 
 

 

0 2 4 6 8 10 12 14

Support Vector Machines

Baysian regularisation

Back Propagation

Feed forward

Standard NN

Fuzzy system

Multi layer perceptron

Learning Vector quantization

Self Organisation Map algorithm

Credit Evaluation, Financial Distress and Other applications


