Predicting U.S. economic recessions and prosperity: a comparative study of machine learning models

The economic cycles, whether it is experiencing recessions or prosperity, serve as the foundation for numerous political and economic decisions. As a result, predicting economic cycles both domestically and globally presents a significant challenge for investors and economic stakeholders. In this study, we investigate the effectiveness of various machine learning (ML) models in predicting economic cycles. Firstly, employs two feature selection methods, including mutual information (MI) and analysis of variance (ANOVA), to select important features. Subsequently, classification models such as Gaussian naïve Bayes, logistic regression, support vector machine (SVM), decision tree, multi-layer perceptron (MLP) neural network, Random Forest (RF), AdaBoost, and voting are utilized to predict economic cycles across various timeframes, ranging from one season to four seasons. This study uses data from the United States (U.S.) economy to evaluate the performance of these models. The results demonstrate the superiority of the ANOVA method in feature selection and the high accuracy of Gaussian naïve Bayes, SVM, and voting models in predicting economic cycles, reaching up to 93% accuracy.