<![CDATA[Tropical diseases remain a major public health problem in Indonesia, particularly in regions with limited access to healthcare facilities, leading communities to rely on traditional medicine based on local wisdom. However, the integration of traditional and modern treatment knowledge in intelligent recommendation systems remains limited. This study aimed to develop a tropical disease treatment recommendation system by integrating Bayesian Network (BN) and Association Rule Mining (ARM). Traditional and modern treatment knowledge were collected from scientific literature and expert interviews, validated by medical practitioners and traditional medicine experts, and incorporated into the system. A quantitative and experimental approach was conducted using a dataset of 150 tropical disease cases comprising dengue fever (42 cases), malaria (35), leptospirosis (28), tuberculosis (30), and leprosy (15). The dataset included 47 symptom attributes, 34 traditional treatment attributes, and 12 modern treatment attributes. Bayesian Network was used to model probabilistic relationships among symptoms, diagnoses, and treatments, while the Apriori algorithm in ARM was applied with minimum support and confidence thresholds of 0.3 and 0.7, respectively. Experimental evaluation on 30 testing cases showed that the integrated BN-ARM model achieved 86.7% accuracy and an F1-score of 86.0%, outperforming standalone BN (82.0% accuracy; F1-score 82.5%) and ARM (79.0% accuracy; F1-score 78.8%). The system generated accurate and contextually relevant treatment recommendations by combining local wisdom and modern medical knowledge.]]>
Copyrights © 2026
