Predicting mergers & acquisitions: A machine learning-based approach

We provide empirical evidence on the predictability of Chinese merger and acquisition (M&A) activities by applying the machine learning approach in corporate finance studies to predict enterprises' M&A activities. We build a comprehensive set of 60 explanatory variables from the literature, employ a variety of widely used machine learning models to predict the occurrence of corporate acquisitions, and compare their predictive power with that of the traditional econometric method represented by the logit model. We show that machine learning has significant out-of-sample forecasting performance for takeovers compared to the logit model. In addition, we rank the importance of the variables and find that these important factors have a noticeable impact on the actual results of M&A prediction. Our findings indicate that utilising machine learning techniques to predict corporate takeover activities is effective and economically meaningful.