Predicting M&A targets using news sentiment and topic detection

This paper uses news sentiment and topics to discuss the challenges and opportunities of predicting mergers and acquisition (M&A) targets. We explore the effect of investor sentiment on identifying M&As targets and how company-specific news articles can be used as a source of sentiment and topics to obtain richer information on various corporate events. We propose a framework incorporating news sentiment and topics into the M&A target prediction model, utilising state-of-the-art transformer-based sentiment analysis and topic modelling approaches. We evaluate the textual features' predictive power using a real-world dataset of US and UK target and non-target companies from 2020 to 2021, with several experiments conducted to reveal the contribution of sentiment and thematic focus of news to M&A target prediction. A profit-based objective function is proposed to overcome the inherent class imbalance problem in the dataset. Our findings suggest that news-based prediction models outperform traditional statistical and single machine learning methods, indicating the need for more robust and less prone to overfitting ensemble learning methods. Additionally, our study provides evidence for the positive effect of news-based negative sentiment on the likelihood of M&A. Our research has important implications for investors and analysts who seek to identify investment opportunities.