A Comparative Study of Deep Neural Network and Statistical Models for Stock Price Prediction

An Investment is a present-day pledge of money or an asset in the hope that it will bring future benefits. An investor can invest in fixed income securities, equities, derivatives, gold, or real estate. An investor’s portfolio can contain a mixture of these assets. Financial Portfolio Optimization is a process that maximizes the return and minimizes the risk for an investor. With increasing population and commodities, finance has become a very complex and vast field. Earlier, investment options were minimal; now, with the introduction of the internet, investment opportunities know no bounds. The portfolio is a collection of assets. An ’Asset’ is an entity that is convertible into cash. Assets bring future benefits. There are two types of assets: movable and immovable. Movable assets are stocks, mutual funds, etc. Immovable assets are physical properties, buildings, land, sophisticated machinery, etc. Assets can also be categorized as tangible and non-tangible. Assets such as gold, vehicle which have physical existence are ’tangible,’ Assets such as patents, copyrights, trademarks, bonds, and stocks are ’non-tangible. A bond gives an investor a fixed income equal to an agreed contract. Stock gives an investor a part of the money earned in the form of dividends. Other non-tangible financial assets are the financial index, interest rate, currency, commodities, etc. The stock market is volatile and difficult to predict. The statisticians and machine learning experts have tried to forecast the stock market. This paper compares the prediction capability of both statistical and machine learning models. The Recurrent Neural Network (RNN), Convolution Neural Network(CNN), Long Short Term Memory (LSTM), and Auto-Regressive Integrated Moving Average (ARIMA) models are compared. It is observed that CNN outperforms other models for the given period of study. It is also observed that machine learning models fair better than the statistical model.