<![CDATA[Product performance classification in marketplaces is a challenge in data analysis because it involves heterogeneous numerical and categorical attributes and an uneven class distribution. This study compares the performance of the Naive Bayes (NB) and K-Nearest Neighbor (KNN) algorithms using a data mining approach on 1,200 data points that have undergone pre-processing. The evaluation was conducted using a cross-validation scheme with accuracy, kappa, precision, and recall metrics. The test results showed that NB was significantly superior with an average accuracy of 80.10% ± 0.61%, while KNN only reached 43.85% ± 1.36%. The kappa, precision, and recall values also showed the consistency of NB's superiority in capturing class distribution patterns. These findings confirm that the probabilistic approach is more effective than distance-based methods that are sensitive to data overlap and feature distribution. Theoretically, this study confirms the importance of algorithm suitability with data structure characteristics in determining classification performance in the marketplace context.]]>
Copyrights © 2026
