A model that reads an order's details and predicts whether it will arrive on time, so a logistics team can spot the shipments likely to run late. It learns from e-commerce shipping records (warehouse block, shipment mode, customer-care calls, ratings, product cost and importance, prior purchases, discount and weight) and compares four classifiers, k-nearest neighbours, logistic regression, a decision tree and a neural network, on accuracy, precision, recall and ROC-AUC.
