AI-Based Sentiment Analysis for Predicting M&A Outcomes

Abstract

This research investigates the role of artificial intelligence (AI)-based sentiment analysis in predicting merger and acquisition (M&A) outcomes by integrating unstructured textual sentiment with traditional financial variables. Using a dataset of global M&A transactions between 2015 and 2023, sentiment signals were extracted from financial news, press releases, analyst reports, and regulatory filings using a hybrid Bidirectional Encoder Representations from Transformers (BERT) and Long Short-Term Memory (LSTM) framework. Sentiment indices were then combined with financial ratios and deal-specific features within ensemble machine learning models to predict deal success versus failure. The empirical results demonstrate that sentiment exerts a statistically significant influence on M&A outcomes, with higher positive sentiment levels strongly associated with successful deal completion. Ensemble models outperformed individual classifiers, achieving superior predictive accuracy, F1-scores, and AUC-ROC values. Feature importance analysis revealed that sentiment indices carried greater predictive weight than leverage or market-to-book ratios, underscoring the growing significance of behavioral and psychological factors in corporate transactions. Regional heterogeneity was also observed, with North America showing higher sentiment-driven success rates than emerging markets. These findings highlight the value of integrating AI-driven sentiment analytics into M&A forecasting, providing insights for corporate managers, institutional investors, and policymakers. Beyond predictive contributions, the study advances the literature on explainable AI in finance by emphasizing transparency in sentiment-based models. The results demonstrate that AI-powered sentiment analysis is not only an innovative complement but also a transformative factor in improving predictive accuracy and strategic decision-making in M&A activities.