AI–POWERED FORE CASTING ALGORITHMS TO OPTIMIZE LAST MILE DELIVERY

Abstract

Last-mile delivery forecasting is a vital component in preserving an advantage over rivals in the  fiercely competitive world of supply chain and e-commerce operations. In order to save  operating costs, this study presents a unique machine-learning technique created especially for  location-based last-mile delivery forecasting. A variety of machine learning algorithms, such as  regression models, decision trees, and neural networks, are used to uncover subtle patterns in the  data by using historical demand data, location-specific characteristics, and economic indicators.  These models are then compared to traditional forecasting models like SARIMA and ARIMA.  The study uses principal component analysis to handle any problems with high-dimensional  data. After extensive hyperparameter adjustment, the final model is chosen and assessed using a  separate dataset. For easier understanding, a graphic flowchart that summarizes the whole  forecasting process is also included. The suggested method demonstrates machine learning's  enormous potential for enhancing last-mile delivery predictions, opening the door for lower  operating costs and increased overall productivity in supply chain and e-commerce businesses.