Late Delivery Risk Prediction in Supply Chain: Moving from Reactive to Proactive Delay Management

<title>Abstract</title> <p> <italic>APL Logistics has hundreds of global orders, the late delivery of which leads to numerous SLA violations, losses, and customer dissatisfaction. Such delays are usually expensive and ineffective to prevent by reactive measures. This paper presents a solution that uses Machine Learning to predict potential late deliveries before their shipment to proactively mitigate risks. After extracting features from historical order records, including Shipping Pressure Index and Region Congestion Index, the model was trained on a dataset addressed with SMOTE technology to counterbalance the imbalance between classes. As a result, the Random Forest classifier reached 72% accuracy, successfully distinguishing three categories of probability: Low, Medium, and High. The model is implemented as a Streamlit dashboard that allows the operations team to redirect or re-prioritize at-risk deliveries and reduce costs by communicating with customers.</italic> </p>