Name of QuantLet : Insider trading Financial Forensics

Published in : Not yet published 

Description : 'Performs dimensionality reduction, visualization, and predictive modeling on 710 features from the insider trading prediction dataset (2014–2019). We apply PCA combined with t-SNE to explore the structure of high-dimensional data and visualize embeddings. In addition, we implement Logistic Regression, Random Forest, XGBoost, and Generalized Random Forest (GRF) to evaluate predictive performance under severe class imbalance, and construct GRF-based confidence bands to analyze heterogeneity and estimation uncertainty.'

Keywords : 
- insider trading
- dimensionality reduction
- PCA
- t-SNE
- logit
- Random Forest
- XGBoost
- Generalized Random Forest
- confidence bands
- SHAP

Author : Yezhou Sha, Wolfgang K. Härdle

Submitted : April 7 2026