Stunting remains a major chronic nutrition challenge in Indonesia, including in the Papua region, with a prevalence far above the national average, thus requiring an explainable yet adaptive early diagnosis tool. This research develops a Smart Decision Support System (SDSS) that hybridizes an expert system based on forward chaining (FC) and certainty factor (CF) with a machine learning (ML) module as a self-learning knowledge base, allowing the measure of belief (MB) and measure of disbelief (MD) values to be periodically recalibrated based on verified case data without losing the explainable nature of the expert rule structure. This system was tested using simulated data from 240 cases with 15 main stunting symptoms and subsequently validated through 5-fold cross-validation and the McNemar significance test. The results of the single split test showed that the static Expert System had an accuracy of 76.39%, which increased to 93.06% after the knowledge base was updated by the ML module, and was higher than the Gradient Boosting model (88.89%); 5-fold cross-validation confirmed the superiority of the hybrid scheme (accuracy 91.25±2.04%) significantly compared to the static Expert System (p=0.016). The largest improvement occurred in precision (from 63.83% to 85.71%) without lowering recall (remained 100%). This study concludes that the integration of expert systems and machine learning through an adaptive mechanism for updating knowledge weights has the potential to produce a more accurate, adaptive, and still explainable clinical decision support system compared to a single approach
Copyrights © 2026