Klasifikasi Status Gizi Ibu Hamil Untuk Deteksi Risiko Dini Stunting Menggunakan Support Vector Machine

Abstract

Stunting is a chronic malnutrition condition that requires early detection during pregnancy (the First 1,000 Days of Life), particularly to address delayed interventions, as evidenced by the 505 recorded cases in Songgon Subdistrict. Given that stunting risk data collection by local public health centers (Puskesmas) is still performed manually, this study develops a computational early detection system using the Support Vector Machine (SVM) method optimized by the Firefly Algorithm. Based on the testing of 400 maternal kohortmedical records, the system identifies Hemoglobin (Hb) levels and Mid-Upper Arm Circumference (MUAC) as the most determinant risk predictors. Evaluation results prove that Firefly optimization is highly effective in handling complex and non-linear data patterns, with the highest predictive performance achieved by the Radial Basis Function (RBF) kernel. This model demonstrates a significant accuracy leap from a baseline of 89.17% to 94.17% after optimization, accompanied by a precision of 94.29%, recall of 92.50%, and an F1-score of 93.31%, ensuring minimal detection failures (false negatives) in high-risk patients. These results outperform the maximum performances of the Polynomial (86.67%), Linear (76.67%), and Sigmoid (65.00%) kernels. Consequently, the hybridization of SVM and the Firefly Algorithm proves capable of producing a robust classification model, making it highly recommended for implementation as a digital early warning system for maternal stunting risk.