Development of Work Program Submission System Using Predictive Data Analytics Based on Neural Network Algorithms at PT Pos Lhokseumawe

Abstract

The digital transformation within PT. Pos Indonesia KC Lhokseumawe demands a more efficient, secure, and adaptive procurement system. This study aims to develop a web-based work program submission system that integrates predictive analytics and neural network algorithms to enhance procurement efficiency and the accuracy of budget forecasting. The system is built using a React.js-based frontend architecture and a FastAPI-based backend, with MongoDB as the database. The N-BEATS model is implemented for time series-based budget forecasting, while Neural Collaborative Filtering is employed to recommend vendors based on interaction history. Evaluation results demonstrate strong performance, with an R-squared value of 0.9965 for the forecasting model and an F1-score of 73.71% for the recommendation model. This integrated system provides procurement management features, budget forecasting, and tender recommendations, and is expected to improve business process efficiency at PT significantly.