Enhancing Profitability Prediction : Insights from Machine Learning, Artificial Intelligence, and Principal Component Analysis

This study examines the efficacy of machine learning (ML) and artificial intelligence (AI) in predicting profitability, offering valuable insights for investors and managerial decisions. Leveraging the extensive array of data associated with Big Data, ML and AI are compared with regression models and random walk predictions for their proficiency in predicting accuracy. We used principal component analysis (PCA) in conjunction with five ML models (decision tree, AdaBoost-DT, support vector machine, back propagation neural network, and K-nearest neighbors) that incorporate AI. The focus is on listed companies in the China A-share market between 2005 and 2020. The PCA approach aids in constructing a comprehensive profitability index (P-index) based on seven profitability indicators. The out-of-sample accuracy of these models is evaluated using average accuracy and area under the curve (AUC) tests. Our results show notable accuracy exceeding 80% on average for all models, and AUC values above 0.850, with the back propagation neural network model achieving an AUC of 0.905. This evidence substantiates the significant role of ML-based AI approaches in consistently predicting profitability with high accuracy. These ML models, particularly pertinent to China's unique political and economic environments, deliver superior predictions of profitability for the Chinese market