Penerapan Regresi Least Absolute Shrinkage And Selection Operator (LASSO) Untuk Mengidentifikasi Variabel Yang Berpengaruh Terhadap Kejadian Stunting di Indonesia

Abstract

Linear regression analysis is an analytical method that can be used to analyze data and draw meaningful conclusions about the dependence of one variable on another variable. In linear regression analysis there are several assumptions that must be met, namely normal distribution, there is no correlation between errors. There are several obstacles that cause the assumption to be unfulfilled, for example the occurrence of correlations between independent variables (multicollinearity). The analysis in this study uses the Least Absolute Shrinkage And Selection Operator (LASSO) regression method with the Least Angle Regression (LAR) algorithm because the stunting data in Indonesia has multicollinearity problems among the independent variables used. LASSO which can solve the case of multicollinearity in the regression at the same time it is possible to reduce the regression coefficient from the highly correlated independent variable to exactly zero. The LASSO coefficient obtained uses quadratic so that the LAR algorithm is used which is more efficient in LASSO computing. Based on the analysis that has been carried out, it is concluded that the variables of exclusive breastfeeding (X1), protein consumption (X2), DPT-HB exercise (X5), maternal height (X8) and diarrhea (X9) had an effect on stunting in Indonesia in 2018.