Title: Visualizing Machine Learning Models for Enhanced Financial Decision-Making and Risk Management
Authors: Ramakrishna Garine, Priyam Ganguly, Isha Mukherjee
Presented at: 2024 Third International Conference on Trends in Electrical, Electronics, and Computer Engineering (TEECCON)
Slide Summary:
This presentation explores how visualization enhances the interpretability and reliability of machine learning (ML) models, especially in high-stakes environments like finance. The focus is on Tsetlin Machines (TM), an emerging ML algorithm that offers greater transparency through logical clause-based decisions. Key contributions include:
Interpretable AI for Financial Risk: The study introduces visualization tools that clarify how TMs make decisions, aiding trust and compliance in banking and financial sectors.
New Visualization-Driven Method: A novel clause-based approach is developed for stochastic asset weighting, improving portfolio management accuracy and insights.
Experiment on Noisy XOR Dataset: Achieved 100% accuracy using a TM model with 10 clauses per class, showcasing TM’s learning dynamics and hyperparameter sensitivity (s, T).
Performance Insights: Demonstrated how TM interpretability bridges the gap between training and real-world performance, making it suitable for financial forecasting and risk assessment.
Future Scope: Suggests enhancements like locally stochastic clause dropping to prevent overfitting and boost performance stability.
Impact & Relevance:
This research offers a transparent, efficient alternative to traditional neural networks, promoting ethical, explainable AI in finance. It aligns with Ramakrishna Garine’s broader body of work in AI-driven supply chain optimization and financial analytics—contributing to both academic research and industry transformation.