Revision as of 11:03, 11 February 2026

Abstract

The logistics and transportation sectors are struggling with major issues like demand variations, disruptions, and inefficiencies, which ultimately undermine the agility and efficiency of the entire supply chain. Most of the time, traditional forecasting models are not entirely accurate in response to life-changing factors like weather, traffic, and inventory levels. The present research intends to build an AI-powered predictive model that can seamlessly enhance not only demand forecasting and logistics but also by the integration of real-time data. The framework incorporates several Machine Learning (ML) models, which are Light GBM for demand forecasting, Random Forest for disruption prediction, Linear Regression for shipping cost estimation, and Support Vector Regression for delivery time deviation prediction. A thorough dataset containing historical demand, weather conditions, traffic, and stock levels was used for the model’s training and evaluation, and its performance was monitored using MAE, MSE, RMSE, and MAPE metrics. The findings indicate that the suggested framework is a lot better than the existing ones, with Light GBM getting the lowest MAE (0.056), MSE (0.005), RMSE (0.072), and MAPE (0.142). This means that the new system can predict much better than before, thus making it possible for the company to take the right decision at the right time and consequently improving the overall supply chain efficiency. The research paper reveals the future possibilities of AI-based solutions for optimising logistics operations and building supply chain resilience.OPEN ACCESS Received: 20/10/2025 Accepted: 25/12/2025 Published: 03/02/2026

Document