While delivery logistics operations are essential to providing excellent customer service, hazards like damage and delays in transit are a constant challenge. Sustaining the transport supply chain’s dependability and quality requires effective risk management. This study compares machine learning algorithms to identify delayed births using two different datasets: the USAID Medical Dataset and DataCo Global. The study assesses how well several machine learning algorithms, which include classifiers and regressors, forecast delays and identify the variables that contribute to late delivery. The purpose of this research is to shed light on the advantages and disadvantages of various machine learning techniques for handling the difficulties involved in late delivery detection. We have discovered that the extra tree approach yields the best results on both datasets after doing a thorough comparative analysis. Additionally, our research indicates that extra tree outperforms the neural network model since its RMSE score is lower for both datasets than the multilayer perceptron. The study’s conclusions can help supply chain managers lessen the possibility of goods being delayed.