VALIDATOR: Video Analytics for Low-Income Earners Using Deep Analytics for Trust Outcome Resolutions

This paper presents VALIDATOR, a novel deep learning-based video analytics system designed to augment credit risk evaluation through the analysis of facial micro expressions. Traditional financial scoring models often marginalize low-income individuals due to limited historical financial data. To address this gap, VALIDATOR integrates a Convolutional Neural Network (CNN) trained on a hybrid dataset to identify combined emotions. The system introduces the Advanced Facial Microexpression Identification (AFMEI) algorithm, which incorporates an Iterative Optimization Strategy for dynamic weight tuning and learning rate adaptation. This enables robust classification of customers as credit-approved or non-approved based on trustworthiness cues captured in real-time facial expressions. Experimental results demonstrate a significant improvement in prediction accuracy over conventional models, highlighting the potential of behaviour-based video analytics in financial inclusion and enterprise risk management.