Abstract

The rapid growth of e-commerce and on-demand food delivery platforms such as Uber Eats, DoorDash, and Meituan has significantly influenced consumer purchasing behavior. To meet rising demand, platforms often rely on either in-house or crowdsourced couriers. While crowdsourcing helps reduce logistics costs, it also introduces operational challenges, particularly around courier performance and behavior, factors that can directly impact delivery efficiency and customer satisfaction. With order volumes increasing at a fast pace, there is a growing need for data-driven strategies that can support better planning and resource management.

This study uses real-world data from the Meituan food delivery platform to conduct an exploratory data analysis (EDA) of courier behavior and performance. Key performance indicators (KPIs) examined include delivery time, distance traveled, courier workload, order acceptance rate, courier activity and inactive time, spatial-temporal delivery patterns, fulfillment rates, and average delivery time during peak and off-peak periods. In addition, several machine learning models: Linear Regression, Random Forest, XGBoost, LightGBM, K-Nearest Neighbors, and Support Vector Machine, are implemented to predict order volumes across different times and regions. These models are evaluated using standard error metrics, including RMSE, MSE, MAE, MAPE and R-squared. By integrating insights from both the EDA and predictive modelling, this study proposes data-driven strategies to enhance operational planning and efficiency.

Keywords: Machine Learning, On-demand Food Delivery, Crowdsourced couriers, EDA, Courier Performance, Spatial-temporal analysis