Authors: Gopichand Talluri
Abstract: Under uncertain conditions, financial decision-making needs methods that combine predictive accuracy, risk management, regulatory requirements, and explainability. Existing financial models, which include both statistical forecasting and conventional machine learning techniques, may fail in terms of describing complex nonlinear market dynamics, quantifying uncertainty in financial decisions, and explaining the effects of financial constraints on decisions. This research presents an ENSC-FI approach to generating financial decisions using a framework consisting of neural learning, symbolic financial rules, uncertainty quantification, and constraint-aware optimization. ENSC-FI seeks to create clear financial decisions using market indicators, risk constraints, budget limitations, and explanation of decisions made. Performance of the proposed framework may be compared to existing traditional machine learning and optimization techniques in terms of return, Sharpe ratio, maximum drawdown, constraint satisfaction rate, and explanation quality. As for an evaluation target of ENSC-FI, the improvement of constraint satisfaction up to 95% and decrease in decision errors by 15% compared to one of the selected baselines will be considered. Those are the targets to be achieved but not actual experimental results yet.
