Deep Bidirectional Temporal Modeling with Boosted Rule Optimization for Dynamic Ride Demand Forecasting in Smart Cities

Abstract

Driver attrition has become a major challenge for ride-hailing and transportation-based organizations, as it directly impacts operational performance, workforce stability, and customer satisfaction. Early prediction of employee attrition enables organizations to implement proactive retention strategies and reduce workforce turnover. Traditional attrition prediction methods mainly relied on statistical analysis, manual evaluation, and conventional machine learning techniques, which often struggled to capture hidden relationships and complex feature interactions within high-dimensional employee datasets. To address these limitations, the proposed system introduces a hybrid deep learning and ensemble learning framework for intelligent driver attrition prediction. The proposed BiLSTM-BRC model utilizes a Bidirectional Long Short-Term Memory (BiLSTM) network to perform advanced feature extraction from structured employee data by learning forward and backward feature dependencies. The extracted deep features are then processed using a Boosted Random Committee (BRC) ensemble classifier consisting of Random Forest (RF), Gradient Boosting (GB), and Decision Tree (DT) classifiers combined through a voting mechanism for robust classification. In addition to the extension model, several comparative machine learning algorithms including Random Forest Classifier, Gradient Boosting Classifier, Support Vector Classifier (SVC), and the Greedy LSTM Tree model are also implemented and evaluated for performance analysis. Experimental results demonstrate that the proposed BiLSTM-BRC extension model achieves superior classification performance with approximately 99% prediction accuracy, outperforming existing baseline and hybrid models on the employee attrition dataset. The complete system is implemented using the Flask web framework with integrated modules for user authentication, model training, prediction analysis, and performance visualization, providing an interactive and efficient web-based HR analytics platform.