The Effect of the Coronavirus Pandemic on the Prediction Accuracy of Stock Price

Abstract

In late 2019, the coronavirus began to spread around the world and impact international politics and economies significantly. In the face of the pandemic, stock markets around the world fluctuated sharply. The study aims to investigate the impact of the pandemic on the predictive variables of a stock prediction model, formed using chip-based variables and sentiment variables derived from comments posted on a social media platform. This study first performs feature engineering analysis to identify the indicators suitable for constructing the prediction model. The analysis then establishes a set of phrase rules to assign sentiment scores to the opinions expressed in replies and evaluates the effect on the accuracy of predictions. The results show that the major chip-based indicators affecting changes in the stock market differ before and after the pandemic. Hence, prediction models should be established separately for analysis in either period. In addition, the results indicate that the model relying on reply-based sentiment scores as a predictive variable provides more accurate predictions of stock price change.