Machine Learning-Based Delivery Time Prediction for Low-Carbon Food Delivery
The increasing demand for online food delivery services has significantly impacted urban logistics, raising critical challenges related to operational efficiency and environmental sustainability. In this context, accurate delivery time prediction plays a key role in improving service quality while supporting the transition toward low-carbon delivery systems. This paper first provides a comprehensive analysis of sustainable and low-carbon delivery modes for urban food distribution. The study then develops and evaluates several machine learning models, namely XGBoost, LightGBM, Gradient Boosting, and KNN, using a dataset that combines operational, temporal, and contextual features to predict food delivery times. The comparative analysis shows that tree-based approaches outperform the other models, with XGBoost achieving the best predictive performance. The findings demonstrate the strong potential of artificial intelligence and machine learning techniques for enhancing delivery efficiency and supporting sustainable urban logistics decision-making. Moreover, the proposed predictive framework can be integrated into optimization models to enhance delivery planning and support the selection of environmentally sustainable delivery modes.