Machine learning techniques via ensemble approaches in stock exchange index prediction: Systematic review and bibliometric analysis

Abstract

The prediction of time series indices in stock markets has become an area of growing interest, providing investors with an advanced insight into a country’s economic prospects in the short, medium, and long term. Much of this type of prediction has been conducted using machine learning techniques, aiming to develop effective and efficient models that can provide greater accuracy in forecasting. In this context, however, single-predict model approaches have shown limitations, such as the inability to handle diverse information efficiently, the non-linearity, and chaos in financial data. In the face of these limitations, ensemble approaches emerge as a strategy to mitigate these challenges. Although these ensemble approaches can be promising, there is no structured understanding related to their application in activities to predict time series indices in stock markets. Within this context, this study aims to investigate to what extent the literature has proposed and evaluated solutions that contribute to characterizing the state of the art in predicting the Stock Market Index through ensemble machine learning approaches. Fifty-three articles resulting from a systematic review, focused on scientific production between 2017 and the first half of 2023, were analyzed. The study performs a bibliometric analysis and categorization of the results, detailing and highlighting the main themes in the field. The bibliometric analysis identifies key authors, countries, conferences, and approaches to predict the Stock Market Index. The qualitative analysis reveals the benefits arising from the combination of algorithms in mitigating accuracy problems in index prediction, as well as systematizing information to characterize the state of the art in the field of study. Our results show that combined algorithms are an effective strategy to increase accuracy in predicting the Stock Market Index, which can benefit investors and financial institutions. This paper presents the future research scope by highlighting gaps in the existing literature for developing techniques to handle concept drifts.