The Use of Machine Learning to Forecast Financial Performance: A Literature Review

Abstract

The paper offers a comprehensive analysis of ten studies covering different facets of the application of artificial intelligence (AI) techniques for identifying financial performance. The financial stability of organizations is a major concern for decision-makers, particularly in the finance field. Diagnosing financial problems in the early stages can prevent further complications. Many of the previous papers have proved the reliability of machine learning in the prediction of financial performance. Therefore, the motivation of this systematic review is to find out how reliable is machine-learning in forecasting financial performance by exploring the pitfalls of machine-learning methods. Examining the models’ accuracies is not sufficient in determining the robustness of the methods applied, however, the harmony and quality of data used are examined as well. Financial performance is categorized as Bankruptcy and Insolvency. The financial datasets related to the study pertain to bankruptcy, data imbalance, feature dimensionality, forecasting insolvency, preprocessing issues, nonfinancial indicators, commonly used machine learning techniques, and performance metrics. Dealing with high dimensionality was suggested by feature extraction and feature selection. Whereas, data imbalance may be prevented by several techniques such as random sampling. The study's conclusions demonstrated the value of dimensionality reduction methods and data balance in data preprocessing. The study illustrates how critical and impactful when taking into consideration the mentioned strategies in enhancing the existent models. The scientific outcome of this work revolves around conceptualizing the cornerstone for building efficient models in predicting financial performance leading researchers to locate unexplored new research avenues.