Predicting Supply Chain Disruptions Using Neural Network Techniques

The supply chain represents the entire product life cycle, from raw materials to the final product's delivery to the customer. Delivery delays are a significant risk that can affect both business operations and customer satisfaction. Such delays can lead to financial losses and diminished customer trust in the service provided. This study aims to propose a model for predicting late deliveries in the event of supply chain disruptions. A multilayer perceptron (MLP) neural network has been used to predict late deliveries for both multi-class and binary classifications, with Support Vector Machine (SVM) and Logistic Regression serving as benchmarks. The models were compared based on accuracy, precision, recall, and F1 score in the provided dataset and the study concludes that the hyper-parameters of the MLP model, such as hidden layers and learning rate, significantly affect its accuracy. The key findings of this research indicates that the tuning hyper parameters and learning rate resulted in varying outcomes and found that MLP model outperforms for binary classification, while in multi-class classification, the MLP, Logistic Regression, and SVM models produced similar results. However, the MLP model achieved the highest accuracy for binary classification.