PCR DAN PLSR ALGORITMA NIPALS DALAM MENANGANI MULTIKOLINIERITAS PADA PREVALENSI STUNTING DI NUSA TENGGARA TIMUR

Abstract

Nutritional problems contribute to 50% of deaths among children under five, particularly in low- and middle-income countries. One of the most common issues in Indonesia is stunting, a condition where a child's height falls below the standard for their age. In 2022, East Nusa Tenggara (NTT) recorded the highest stunting prevalence in Indonesia at 35.3%. However, quantitative statistical analyses of its contributing factors in NTT remain limited. This study aims to compare partial least squares regression (PLSR) using the NIPALS algorithm with principal component regression (PCR) in addressing multicollinearity. The secondary data were obtained from the 2022 Indonesian Nutrition Status Survey (SSGI), published by the Ministry of Health and BPS NTT, consisting of one response variable and ten predictor variables. Results showed that the PLSR model outperforms PCR, with an adjusted R² of 0.741 compared to 0.322. The superiority of PLSR is also evident from its lower RMSE and MAE values (2.783 and 1.910) compared to PCR (4.742 and 3.346). PLSR identified five significant predictors: average daily protein consumption per capita, number of children receiving DPT and HB immunizations, Human Development Index, percentage of households with access to safe drinking water, and number of people living in poverty.