Influence factor studies based on ensemble learning on the innovation performance of technology mergers and acquisitions

Abstract

Technology merger and acquisition (M&A) is an important way for companies to enhance their innovation capabilities, and it is important to clarify the factors that affect the innovation performance of technology M&A. Existing studies only focus on the association between individual factors and technology M&A innovation performance, lack comparative analysis of different dimensions. Additionally, most studies discuss within-sample prediction results, whose findings may not apply to the full sample. Based on ensemble learning approaches in machine learning, we discuss the differences in the predictive power of multidimensional influence factors on the technology M&A innovation performance and then identify the main factors and the variables with the strongest predictive power. It is found that: (i) the M&A motivation of the acquirer is the most significant factor affecting the innovation performance of technology M&A; (ii) AdaBoost has the strongest explanatory power and the highest predictive accuracy among traditional machine learning approaches; (iii) among the multidimensional feature variables, the total asset turnover, overhead rate, executive compensation ratio, industry-wide misvaluation, average R&D investment ratio, average R&D staff number ratio, corporate risk-taking level, analyst attention, and media attention have the best predictive effect on technology M&A innovation performance. Ensemble learning approaches have a better out-of-sample generalization and can identify key influencing factors by comparing multiple dimensions in the performance of technology M&A innovation prediction.