- ANALISIS FAKTOR-FAKTOR YANG MEMENGARUHI DIABETES MELITUS DI JAWA BARAT MENGGUNAKAN MULTIVARIATE ADAPTIVE REGRESSION SPLINES (MARS)

-

Abstract

Diabetes mellitus is a metabolic disorder characterized by chronically elevated blood sugar levels due to impaired insulin secretion or action. West Java ranks second among Indonesian provinces with the highest number of diabetes mellitus cases based on medical diagnoses. In 2022, the number of diabetes mellitus cases in this province reached its lowest point in the past five years, suggesting the influence of significant factors contributing to this decline. Therefore, identifying the factors affecting diabetes mellitus prevalence in West Java in 2022 is essential. However, the data exhibit no clear pattern, necessitating a nonparametric regression approach for modeling these factors. This study employs Multivariate Adaptive Regression Splines (MARS), a flexible method that partitions data into segments and applies linear regression within each segment. Model selection criteria include Generalized Cross Validation (GCV) and Akaike Information Criterion (AIC). Based on GCV, seven significant variables were identified, whereas AIC indicated eight significant variables influencing diabetes mellitus prevalence in West Java. Structural differences between models are also observed in the number of basis functions: the GCV model utilizes 13 basis functions, while the AIC model employs 14. In terms of model performance, the GCV model achieves an R² value of 0.994, whereas the AIC model attains an R² value of 0.995.