Article Preview
Top1. Introduction
The stock market is an aggregation of sellers and buyers and serves as a platform for exchanging different companies' stocks. A group of companies constitutes a dedicated stock exchange index aggregated country-wise. Digitization of the stock exchange has facilitated algorithmic trading (AT) on state-of-the-art-machines. According to a World Economic Forum report, over 50% of the trades are on AT. Investors can now complete stock market transactions rapidly.
In this context, researchers are attempting to measure the implications of trading with modern techniques and tools like Artificial Intelligence and robotics in a market where human biases dominate. According to a report of the World Economic Forum, the global market is witnessing a new trend. Presently, 10% of trades are performed by retail traders without the involvement of AT, 40% trades are based on decisions to invest in stock market/index funds/ETFs, and the balance 50% trades are conducted on automatic trading using computers, also known as algorithmic trading (AT). Prediction of the futures markets in the context of AT is an exciting area for further research.
To predict stock market trends, we need to identify the determinants that cause daily stock price changes. One set of factors that reflect the movement in the stock market is technical indicators. These are leading indicators of variation in stock market trends since they perform mathematical transformations on historical stock prices.
In the context of algorithmic trading, many technological advances have taken place, leading to the development of robust algorithms. We implement these algorithms to determine whether the stock market follows an upward trend (bull phase) or tanks (bear phase).
The model enables real-time decision-making of whether to buy or sell or hold onto the stocks. Machine learning techniques are imperative to adopt in real-time decisions with a high level of confidence and accuracy. In machine learning (Demir et al., 2020; Gan et al., 2020), the computer systems are trained on previous instances (also called ‘train set’) and implicitly programmed to respond to similar real-time scenarios (test set) for decision-making. The machine learning algorithms consider the historical stock prices and leading technical indicators as factors and train the model for future trend predictions (Manickavasagam, Visalakshmi & Apergis, 2020).
The paper aims to develop a stock price prediction model from the above factors training using Support Vector Machine, Random Forest, Gradient Boosting, and Deep Neural Networks for the Indian stock market index (NSE NIFTY 50). We compare the algorithms' predictive performance and compute the variable importance to identify the most critical factors' index trend. The rest of the paper is structured as follows.
Section 2 reviews the literature on technical analysis, technical indicators, and machine learning techniques. The data and research methodology is in Section 3. Section 4 discusses the results of classifier performance and variable importance. Finally, in Section 5, we discuss the conclusions of the research followed by references.