Smart Logistics Using IoT: A Data-Driven Approach to Reduce Transportation Delays

Urban last mile delivery is a key part of smart city logistics, but one that is hampered by delays and has a major impact on service reliability and operational efficiency - and sustainability. Despite the progress made in Internet of Things (IoT) technologies and artificial intelligence, transportation delays are still rampant, as a result of urban congestion, spatial heterogeneity and demand volatility. This study examines how big data in the logistics industry, made available through IoT, can be harnessed using machine learning algorithms to predict and minimize last mile delivery delays. Using the LaDe dataset that contains more than 4.4 million real-world delivery records across five major Chinese cities, the study combines a quantitative data-driven method combining exploratory analysis, supervised learning, hyperparameters optimization, and simulation-based intervention evaluation. Logistic Regression, Random Forest and LightGBM models were created and tested, with a random forest model fine-tuned showing the best results, (80% accuracy and 0.857 ROC-AUC). Spatial and temporal features proved to be the most influential predictors, and thus the importance of local urban context. A risk-based simulation showed that targeted interventions for higher-risk deliveries are likely to enable a reduction of overall delay rates by over 8% with city-specific improvements of more than 20% in heavily congested areas. The findings confirm the value of operationalizing predictive analytics to provide meaningful reductions in delays and enable proactive and data-driven decision making in smart urban logistics systems.