A Comparison of Machine Learning Algorithms to Predict Stock Price During COVID-19 Outbreaks: An Empirical Investigation of Pakistan Stock Exchange

Abstract

The research investigates the performance of four machine learning algorithms ‘Linear Regression, Support Vector Machine, Gradient Boost and Random Forest’ in predicting stock prices in Pakistan during the COVID-19 pandemic. Utilizing 15 trading factors sourced from the Pakistan Stock Exchange website and analyzed using Python programming language, the study calculates volatility during the pandemic period. Results reveal that the Random Forest model consistently exhibits robust predictive capabilities across various metrics, outperforming other algorithms by accurately predicting stock prices based on factors such as the previous day’s closing price, open price, average price, turnover and P/E ratio, with over 90% accuracy. Notably, Random Forest also surpasses in predicting the lowest price, highest price, P/B ratio, P/S ratio, fluctuations in currency (in Rupees), turnover rate, total market capitalization and total equity. This study contributes to the literature on stock return algorithms, uniquely examining the effectiveness of different financial factors with various machine learning algorithms in the context of Pakistan’s stock market during a challenging period. The findings hold significance for investors, traders and policymakers, offering insight for decision-making and regulation formulation amid market instability.