<![CDATA[Purpose – This study compares five machine learning algorithms for predicting supply and sales categories in the Honda oil spare parts supply chain in Banten. Using an exploratory data mining approach, the models were evaluated through confusion matrix analysis and optimized using hyperparameter tuning to improve classification accuracy and model performance. Methods – This study uses a data mining approach with data selection, preprocessing, transformation, modeling, and evaluation. Model evaluation was performed using a confusion matrix with accuracy, precision, recall, and F1-score metrics. Model optimization was implemented using one of the hyperparameter tuning techniques, that is random search cross validation. Findings – Random Forest and XGBoost achieved identical and the highest performance after hyperparameter tuning, with an accuracy of 0.8061. Decision Tree showed a very close performance with an accuracy of 0.8060. Neural Network achieved an accuracy of 0.7939, while Naïve Bayes recorded the lowest performance at 0.5515. Overall, hyperparameter optimization improved the performance of most models, demonstrating its effectiveness in enhancing classification accuracy across different algorithms. Implications – This study provides a systematic machine learning framework for supply category classification and demonstrates the practical potential of machine learning and hyperparameter tuning in analyzing Honda motorcycle oil spare part supply patterns using historical operational data. Originality – This study presents a structured approach that compares various algorithms but also integrates confusion matrix-based evaluation and hyperparamter tuning optimization. Nevertheless, the optimised model requires further validation using a cost- and operationally-sensitive evaluation approach to assess the actual business impact under real-world supply chain conditions.]]>
