This research investigates the application of Operations Research techniques, specifically heuristic algorithms, for optimal supplier selection, a critical multi-criteria decision-making problem in supply chain management. Effective supplier selection significantly impacts an organization's financial resources and overall success, necessitating consideration of both qualitative and quantitative factors. The study proposes a novel heuristic approach to predict the most cost-efficient supplier by predicting "Line-Item Value" using supplier and order features. Two machine learning models, Ridge Regression and Support Vector Regression (SVR) with an RBF kernel, were employed and evaluated using Root Mean Squared Error (RMSE). The methodology involved rigorous data preparation, including filtering non-numeric values and removing low-variance features, followed by hyperparameter tuning using scikit-learn's Grid Search with Leave-One-Group-Out cross-validation and an iterative hill climb approach. The results demonstrate the successful training of both models to predict cost-efficient suppliers, with optimal parameters identified through Grid Search. For Ridge Regression, an optimal alpha of 0.85 yielded an RMSE of 0.08, while SVR achieved an RMSE of 0.08 with a C value of 0.89. These findings underscore the effectiveness of heuristic algorithms in addressing complex supplier selection problems and offer improved performance and efficiency compared to traditional optimization techniques.

Keywords

Operations research techniques, Heuristic algorithms, Optimization, Ridge regression, Support vector regression