Applications of Machine Learning to predict outcome of Mergers and Acquisitions

Specialeforsvar: Cian Thomas Stanton

Titel: Applications of Machine Learning to predict outcome of Mergers and Acquisitions

Abstract: This thesis aims to investigate the efficacy of machine and deep learning models in predicting the success of an announced merger and acquisition using a wide range of features relating to the deal structure, financial variables, and macroeconomic factors, among others. The deal space analysed is those with public targets and acquirers between Jan 2003 and Aug 2023. The models are then applied to some basic merger arbitrage strategies, following tuning, to test the potential to generate excess profits by either going long or short on the target. The results suggest that the most important features relate to the target’s attitude in the deal. The models were very similar regarding the accuracy of predictions but XGBoost and Support Vector Machine with RBF kernel performed exceptionally well compared to the rest. The results of the trading analysis suggest
that there is strong potential for a strategy to be developed related to the shorting of the target company when a model is trained and tuned to maximise the precision.

Vejleder: Jostein Paulsen
Censor:   Mette Havning