Predicting Supply Chain Risks Using Support Vector Regression (SVR) And K-Means

Abstract

Supply chain management is very important in maintaining a smooth flow of goods and services in the supply chain. However, there are numerous risks to chain disruptions, such as weather or economic uncertainty. Emerging risks are not adequately known by traditional risk management approaches. The current paper offers a unified framework of machine learning. It incorporates the (SVR) model for predicting risk level in conjunction with K-means clustering to classify risks in the supply chain. The model makes predictions from data recorded over 10,000 times from various industries (manufacturing, retail, logistics, etc.) The SVR model is optimized using GridSearchCV, while the number of clusters used in the K-means model is based on the elbow method. The performance of the IT model is assessed using several key performance indicators (Accuracy, F1-Score, RMSE, and MSE). The combined SVR + K-means method performed better than other baseline models, including linear regression (Accuracy = 78%, F1-Score = 0.72) and the individual K-means technique (Accuracy = 82%, F1-Score = 0.80). The results show that by using the combination of both techniques, an accuracy of 88% and an F1-Score of 0.91 were attained. RMSE for the model was 0.32. The results demonstrate the advantages of combining regression and clustering techniques to enhance decision support and resilience in the supply chains, which can better address risks. The model could be further improved and tested in other industries in the future.