Copyright © 2018 Avilasa Mohapatra et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. International Journal of Engineering & Technology, 7 (2.6) (2018) 71-76 International Journal of Engineering & Technology Website: www.sciencepubco.com/index.php/IJET Review Paper Applications of neural network based methods on stock market prediction: survey Avilasa Mohapatra 1 *, Smruti Rekha Das 1 , Kaberi Das 1 , Debahuti Mishra 1 1 Department of Computer Science and Engineering, Siksha ‘O’ Anusandhan Deemed to be University, Bhubaneswar, Odisha, INDIA *Corresponding author E-mail: avilasa29@gmail.com Abstract Financial forecasting is one of the domineering fields of research, where investor’s money is at stake due to the rise or fal l of the stock prices which unpredictable and fluctuating. Basically as the demand for stock markets has been rising at an unprecedented rate so its prediction becomes all the more exciting and challenging. Prediction of the forthcoming stock prices mostly Artificial Neural Network (ANN) based models are taken into account. The other models such as Bio-inspired Computing, Fuzzy network model etc., considering statistical measures, technical indicators and fundamental indicators are also explored by the researchers in the field of financial application. Ann’s development has led the investors for hoping the best prediction because networks included great capability of machine learning such as classification and prediction. Most optimization techniques are being used for training the weights of prediction models. Currently, various models of ANN-based stock price prediction have been presented and successfully being carried to many fields of Financial Engi- neering. This survey aims to study the mostly used ANN and related representations on Stock Market Prediction and make a proportional analysis between them. Keywords: ANN; Financial Forecasting; Stock Market Prediction. 1. Introduction Stock market involves a zone of open interest where shares are bought and sold by investors and company respectively. Here the investors include the public for the purpose of rising of the capital and a bond of security is involved in between the two. The im- portance of stock market can be analyzed from the fact that it pro- vides a ground for raising money with the help of stock shares and commercial bonds. With the help of this, even the investors are ben- efited as they are given the shares of company’s profit. As a finan- cial barometer, stock market has influenced every strata of society, from the average family to the wealthiest. Stock market prediction involves the pre forecasting of the probable share prices based on past share prices analysis. Some of the factors which tends to affect the stock market prices are the internal developments within the company, world events (such as world war, natural disaster, and terrorism), inflation and interest rate, exchange rates with foreign currencies, and the last one is advertisement or hype of new event or product which results in promotional event. Stock market analysis is gaining grounds of popularity may be due to its volatility, fluctuating and inconsistent nature, for which it be- comes difficult for investors to preplan their strategy of investment in share market. Due to which it becomes one of the important rea- sons for prediction of future possible values of stock prices. The volatile nature of stock market here refers to the instabilities of the market value for returns of the shares held by the investors. Being volatile in nature it becomes difficult for accuracy and better guid- ance of investors. Due to its volatility, the result predicted are not 100% accurate but it helps investors to anticipate the future. Stock prediction should be as accurate as possible, so that it can be help- ful, due to which technical analysis is performed prior to prediction. Analysis of previous historical data helps in getting an overview of changes in day to day basis and creating a statistical graph. Another reason for increase in the demand of stock market prediction is its challenge of being dynamic and possesses many hypothetical and experimental limitations. Any predictive model can be broadly cat- egorized into following categories: non-parametric, parametric and semi-parametric model, where semi-parametric model includes fea- tures of both parametric and non-parametric model. Predictive modeling is a statistical based concept. The model may use a com- plex neural network or maybe a simple statistical formula. Some of the mostly used prediction tools in previous works are ANN (Arti- ficial Neural Network), FLANN (Functional Link Artificial Neural Network), BPNN (Back Propagation Neural Network) and some others are RBFNN (Radial Basis Function Neural Network), ELM (Extreme Learning Machine), etc. Here, the above techniques’ ap- plication in stock market prediction is analyzed. All the above tech- niques have their own pros and cons. The practical applications of some of the techniques on stock market analysis are proposed. Here are overviews to some of the above methods with a brief descrip- tion. ANN is an interconnected group of nodes, which functions similarly to that of a human brain and is very much preferred due to its superiority and variations [1]. ANN contains several other mod- els, for example, MLP (Multi-Layer Perceptron Network), Autoen- coder etc. Hadavandi et al. [2] anticipated a ANN model based on genetic fuzzy system for predicting the next day’s stock prices as well as overcoming the shortcomings of ANN model alone, as every Artificial Intelligence (AI) model has its own pros and cons. BPNN is one of the techniques which are most often used for stock market prediction due to its complexity and amalgamated multi- layered neural network [3] and here the author proposes a modified BPNN technique overcoming some of its previous drawbacks. Zhang et al. [4] proposed a hybridized model based on EMD- BPNN, where EMD (Empirical Mode Decomposition) is used to