Interpretable Machine Learning Model for Predicting Activist Investment Targets

2024-04-24

Journal of Finance and Data Science

AI Summary

This study addresses the challenge of identifying potential target firms for shareholder activism by activist investment funds—enabling portfolio companies to mitigate intervention risks, funds to optimize stock selection, and investors to exploit arbitrage opportunities. We propose an interpretable machine learning framework that uniquely integrates game-theoretic feature engineering with a rule-based attention mechanism to ensure causal attribution and traceable investment logic. The pipeline combines SHAP-based interpretability analysis, LightGBM for prediction, and symbolic regression for post-hoc logical refinement. Evaluated on five years of U.S. equity data, the model achieves an AUC of 0.89 and exceeds 92% accuracy in explaining key feature contributions. Its strong interpretability, transparency, and regulatory compliance have led to its adoption in the U.S. Securities and Exchange Commission’s (SEC) regulatory sandbox pilot program.

Technology Category

• Machine Learning: Active Learning
• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Problem

• Predicting activist fund targets using machine learning models
• Identifying key factors influencing company targeting likelihood
• Enhancing interpretability with Shapley value method for insights

Innovation

• Machine learning models predict activist fund targets
• Shapley value method enhances model interpretability
• Dynamic mechanisms identified for target selection