EXPLAINING A STOCK RETURN VOLATILITY PREDICTION MODEL USING A MACHINE LEARNING APPROACH

Abstract

The present study aims to develop a machine learning model for predicting stock return volatility. This research is applied in purpose and quantitative-analytical in method. The statistical population includes all companies listed on the Tehran Stock Exchange and Securities Market. The required variables in this study consisted of end-of-day stock price data (opening price, highest price, lowest price, average price, and volume) collected over the required time period. Using fuzzy neural networks and Markov methods, the intended data were analyzed. For this purpose, MATLAB software was used. Libraries such as Pandas, NumPy, and Scikit-learn were used to generate noise-free data and construct the model. To develop an algorithm for predicting stock returns, machine learning models were employed. For the initial model, a linear regression was built in Python using scikit-learn to predict stock returns. The results showed that, after accounting for transaction costs, the machine learning classifiers used in this study indicate that Support Vector Regression cannot generate positive Sharpe ratios in all cases. In contrast, logistic regression and multilayer perceptron can achieve them in some cases. Trading strategies based on a designed four-day stochastic prediction can generate positive returns when total transaction costs are below 0.015–0.040 per transaction