Analysis of the performance of predictive models during Covid-19 and the Russian-Ukrainian war

Vancsura, László and Bareith, Tibor (2023) Analysis of the performance of predictive models during Covid-19 and the Russian-Ukrainian war. PÉNZÜGYI SZEMLE/PUBLIC FINANCE QUARTERLY (1963-), 69 (2). pp. 118-132. ISSN 0031-496X

Preview

Text 1399-Cikkszovege-7018-1-10-20230630.pdf Download (361kB) | Preview

Abstract

In our paper, we investigate how effectively artificial intelligence can be used to predict stock market trends in the world’s leading equity markets over the period 01/01/2010 to 09/16/2022. Covid-19 and the Russian-Ukrainian war have had a strong impact on the capital markets and therefore the study was conducted in a highly volatile environment. The analysis was performed on three time intervals, using two machine learning algorithms of different complexity (decision tree, LSTM) and a parametric statistical model (linear regression). The evaluation of the results obtained was based on mean absolute percentage error (MAPE). In our study, we show that predictive models can perform better than linear regression in the period of high volatility. Another important finding is that the predictive models performed better in the post-Russian-Ukrainian war period than after the outbreak of Covid-19. Stock market price forecasting can play an important role in fundamental and technical analysis, can be incorporated into the decision criteria of algorithmic trading, or can be used on its own to automate trading.

Actions (login required)

Edit Item