The application of feed forward neural networks to merger arbitrage: A return-based analysis

Abstract

This study examines the effectiveness and applicability of a trending machine learning algorithm, the feed forward neural networks (FFNNs) in making merger arbitrage investment decisions. Using a sample of attempted takeovers, 24 deal-specific, target-specific, and macroeconomic factors serve as input variables for the proposed FFNNs model. The resulting failure probabilities are utilized by a simulated hedge fund in evaluating merger arbitrage opportunities. By comparing other funds employing simplistic or commonplace predictive models and investment decision rules, our findings reveal the power of machine learning in takeover failure prediction and the use of FFNN can increase risk-standardized deal returns on average.