<![CDATA[Selecting the appropriate drug therapy is a crucial aspect of clinical decision-making to improve healthcare quality. Utilizing data mining on medical records enables the extraction of objective, data-driven therapy patterns. This study aims to compare the performance of the ECLAT and Decision Tree algorithms in generating drug therapy recommendation rules based on patient complaints and diagnoses. The dataset consists of 1,000 multi-label patient visit records. In the ECLAT method, association rules are generated through frequent itemset mining with a minimum support range of 0.01–0.20, while in the Decision Tree method, rules are extracted from the paths of a tree trained using a multi-output model. To ensure a fair and unbiased evaluation, experiments are conducted using 5-fold cross-validation, where rules are generated from training data and recommendation performance is evaluated on unseen test data. Rule characteristics are analyzed using interest measures such as confidence, lift, and Zhang’s metric, while recommendation performance is evaluated using Hit@k, Precision@k, and Recall@k. Experimental results show that ECLAT consistently generates a higher number of rules and stronger association patterns, as indicated by higher average confidence values. However, Decision Tree demonstrates slightly better performance in Precision@3 and Recall@3, indicating more accurate Top-K recommendations. These findings suggest that ECLAT is more suitable for comprehensive pattern discovery, while Decision Tree is more effective for generating precise and actionable recommendations. Both methods exhibit complementary strengths for developing transparent and interpretable clinical decision support systems]]>
