Detecting and Predicting Supply Chain Risks: Fraud and Late Delivery Based on Decision Tree Models

Abstract

In modern supply chains, fraudulent orders and late deliveries cause major disruptions, leading to inefficiencies and increased costs. Traditional methods like manual audits and rule-based systems are often inadequate. They struggle to handle complex data and adapt to rapidly changing conditions. Machine learning provides a more effective solution by managing large datasets and detecting intricate patterns. This study examines decision tree models for detecting and predicting risks within supply chains. This research takes the data smart supply chain dataset as an example, analyzing the effect of deploying a decision tree into risk prevention. After data cleaning and feature engineering, the decision tree analyzes feature importance, helping detect key factors that cause risks. Then, a decision tree model is built to determine whether an order is fraudulent and predict whether it will be delivered late. The model's performance is measured using accuracy, precision, recall, and F1-score. The results show that decision trees are an effective tool for identifying these risks. They offer clear insights into key factors impacting supply chain performance. This study concludes that machine learning can improve risk management in supply chains. It helps make operations more efficient and resilient against disruptions.