The role of M&A motivations in predicting M&A: Fundamental machine learning approach

Abstract

This thesis tested the ability of industry sector M&A motivations to increase the M&A prediction when combined with target-related financial and industry-related variables. The literature review summarized how M&A processes and M&A motivations are defined in the previous M&A literature. Also, previous studies in M&A prediction and their used variables and techniques were evaluated to produce the best possible predictive power. In total five different algorithms were used. M&A motivations were defined with text data mining techniques and extracting motivation keywords from news articles. Based on the literature review benchmark model was constructed and three other models were compared to this. These models were evaluated by their accuracy, precision, recall, F1 scores, and AUC values. The data sample was constructed from Finnish M&A that occurred during 2015-2019 and from news related to M&A during 2014-2019.

M&A motivations increased the predicting ability of machine learning algorithms when compared to benchmark but when combined with industry-related variables the increase in predicting power was negligible. Average results produced by state-of-art machine learning algorithms were better than results produced by logistic regression. The best per-forming algorithm across all used models was the random forest.