Class Actions and Beyond: Unraveling the Link Between Corporate Decision-Making and Risk Prediction in the Age of Machine Learning

This study delves into the domain of corporate litigation risk, particularly through the prism of securities class actions, by leveraging financial datasets and trading metrics to broaden the scope of predictability for securities class-action litigation within the wider context of business risks. We delineate the interconnections between class actions and financial risks, further exploring the capability to forecast these risks as precursors to future class actions. We expand financial risk analysis through a nuanced comparative study, utilizing an array of machine learning methodologies. Our findings reveal that decision trees perform well in-sample, though their out-of-sample performance falls short compared to linear models with nonlinear features. However, logistic models, gradient boosting, and k-nearest neighbors (KNN) models show promising results in enhancing out-of-sample performance, albeit at higher computational costs arising from hyperparameter tuning.