Classification of Vendor Categories Using Machine Learning in Industrial Decision Support Systems

Long-term engagement with vendors is established through a crucial supply chain management process of vendor selection. Vendor selection has a significant effect on a company’s success or failure. Incorrect vendor selection leads to substandard goods or services, project delays, user dissatisfaction, and material or nonmaterial losses. Research on supplier selection using machine learning often overlooks automated feature selection mechanisms, which are essential to improve the evaluation process. These mechanisms concentrate on pertinent features, ensuring precision and minimizing the risk of overfitting. To address this research gap, we developed an automatic vendor evaluation method using Chi-square statistics, Logistic Regression coefficients, and Recursive Feature Elimination (RFE) to identify and exclude less relevant features from the evaluation process. The hyperparameter tuning of the Random Forest model determined the optimal hyperparameter values to enhance the predictive performance while minimizing overfitting risks. The model achieved an overall accuracy of 0.9834. In addition, the scores of precision and recall exceeded 0.97 for the "Good" and "Fair" categories, demonstrating its effectiveness in vendor classification and providing critical insights for improving vendor evaluation processes. This study improves the vendor selection procedure and makes a substantial contribution to the use of artificial intelligence in business decisions. This offers a solid framework for automated assessments that can be tailored to different decision-making scenarios.