DeepStressScan: A Dual-Forest Intelligence Model for Anomaly Detection and Stress Prediction

Abstract

Identifying extreme stress early is crucial for timely mental health intervention. However, traditional predictive models frequently face challenges with outliers and a typical trend in psychological data. This research introduces DeepStressScan, a combined anomaly detection system that integrates Isolation Forest and Random Forest methods to more accurately identify and predict individuals experiencing high stress. The model functions in two phases: initially, Isolation Forest separates unusual stress responses without any prior labels; subsequently, Random Forest utilizes this enhanced dataset to forecast mental health outcomes with greater precision. By integrating unsupervised and supervised learning, DeepStressScan improves anomaly detection while also increasing the dependability of stress outcome forecasts. Studies performed on datasets for mental health evaluation show that this dual-forest method surpasses single-model benchmarks in terms of precision and stability. DeepStressScan provides a scalable and understandable approach to incorporating anomaly detection in mental health analysis, which could be useful for clinical evaluations and digital well-being services.