Diagnosing Stunting Risk in Toddlers Using Mamdani Method

Abstract

Stunting remains a major public health problem that requires early and accurate risk assessment to support effective prevention strategies. This study proposed and evaluated a Mamdani-based fuzzy expert system to assess stunting risk in toddlers by integrating six key determinants: birth length, dietary diversity, protein intake, immunization status, infectious disease history, and breastfeeding practice. The system modeled uncertainty and partial membership through fuzzification, rule-based inference using the minimum operator, and centroid-based defuzzification to generate a quantitative stunting risk score and categorical risk level. We implemented a prototype of the fuzzy expert system and validated its behavior using representative hypothetical cases and a small set of anonymized real data provided by a local health center nutritionist. The fuzzy rule base was first verified to ensure logical consistency and intuitive outputs under extreme input conditions. Experimental results showed that the system correctly classified all evaluated cases in accordance with expert assessments, achieving 100% agreement in risk category assignment. For illustrative cases, the system produced a risk score of 80 (High Risk) for a child with multiple adverse factors and 17.62 (Low Risk) for a child with favorable nutritional and health conditions. These findings demonstrate that the proposed Mamdani fuzzy expert system can effectively handle uncertainty in stunting risk assessment and provide interpretable, nuanced outputs suitable for decision support. The approach shows strong potential for assisting healthcare workers in early screening and prioritization of interventions. Future work will focus on large-scale clinical validation, optimization of membership functions using empirical data, and deployment of the system in web-based or mobile platforms to support practical stunting prevention programs.