MODEL REGRESI MENGGUNAKAN LEAST ABSOLUTE SHRINKAGE AND SELECTION OPERATOR (LASSO) PADA DATA BANYAKNYA PENDERITA GIZI BURUK KABUPATEN/KOTA DI JAWA TENGAH

Abstract

Malnutrition is the most severe form of the occurrence of chronic malnutrition. Malnutrition is influenced by many interrelated factors. In this study, carried out the modeling of the factors that influence malnutrition using Least Absolute Shrinkage and Selection Operator (LASSO) method with Least Angle Regression (LARS) algorithms due to the factors that influence malnutrition there is multicollinearity detected. LASSO shrinks the regression coefficients of the independent variables that have a high correlation to be right at zero or close to it. LASSO coefficients calculated using quadratic programming so that LARS algorithm is used due to efficiency on LASSO computing. Based on the analysis performed, the model of LASSO in malnutrition data at Central Java Regency/City in 2014 was obtained in the second stage when the value s = 0,02 with MSE value of 0,82977. Concluded that the infants variable (0-6 months) which got exclusive breastfeeding, household that behave with clean and healthy life, infants immunized against Hepatitis B, baby immunized against DPT-HB3, house with proper sanitation, and house with drinking water which accordance with health requirements affect the infant malnutrition in Central Java in 2014. Keywords: malnutrition, multicollinearity, LASSO, LARS