LSTM-Based Predictive Analytics for Resource and Cost Optimization in Project Management

Accurate forecasting of resources and effective cost management are essential in project execution. Conventional models often fail to address the dynamic nature of projects with multiple dependencies and uncertainties. This study introduces a predictive framework based on Long Short-Term Memory (LSTM) networks, designed to capture temporal dependencies and sequential patterns in project data. The model integrates data preprocessing, temporal encoding, and bidirectional stacked LSTM layers to forecast task duration, resource allocation, and project delays. Using historical datasets covering schedules, resource allocation, and project risks, the LSTM model significantly outperformed baseline approaches. It achieved a Root Mean Squared Error (RMSE) of 0.05, and R² = 0.96. Results show a 20% reduction in project cost and an improvement in resource utilization from 65% to 85%.