M&A outcome prediction - Supervised Machine Learning
Key words : Merger & Acquisition, Firms Screening, Due Diligence, Value Prediction, Machine Learning, Supervised Learning, Logistic Regression, Support Vector Machine, Decision Trees, Lower Risk, Predictive Analytics.
Abstract : In this paper, we investigate the topic of M&A, and more specifically its predictability using supervised machine learning techniques. M&A transactions are known to be particularly complex, and to mobilise significant human and financial resources, and involve many different actors. It is interesting to study the factors influencing the outcome of these transactions in order to reduce the risk of bad targets for companies wishing to use such solutions to grow. In order to understand the problem as well as possible, we first explain what an M&A deal is and why companies resort to it. The currently achieved work have often used Logit and Probit techniques, which are often inaccurate when dealing with non-linear analyses, such as M&A deals. In order to analyse our dataset, coming from deals in the industrial sector (aerospace, manufacturing...), we have created an algorithm in Python language using the main supervised learning techniques. The results obtained confirm the inadequacies of linear models, and highlight the relative efficiency of decision trees to predict deals successfulness.