Analisis Perbandingan Model Regresi Linier Metode Forward Selection, Backward Elimination, dan Stepwise Regression (Studi Kasus Pendapatan Oli Mesin pada UD. Bengkel Mobil Abadi)

Abstract

In selecting variables for a regression model, independent variables can explain the dependent variable and meet the assumptions of linear regression so that the regression model is BLUE (Best Linear Unbiased Estimator). There are two estimates used, namely Maximum Likelihood and Ordinary Least Square. There are three methods used, namely forward selection, backward elimination, and stepwise regression. These three methods are variable selection where these methods will select the existing independent variables. Selecting this variable is necessary to obtain significant results on the dependent variable. Of the three variable selection methods, the best model will be produced. Next, the results of the three variable selections will be analyzed to obtain different results or not. This research uses five independent variables, namely the vehicle changing the oil (X1), customer (X2), profile interactions (X3), the duration of the oil change (X4), the difference in the distance between vehicles changing the oil (X5), with the dependent variable being engine oil revenue. By using data on trading businesses, we look for variables that have a significant effect on the dependent variable using the three methods. The research results show that these three methods have significant and similar regression model results. The Backward Elimination method is more efficient and takes less time in steps in the variable selection process than Forward Selection and Stepwise Regression. The model result is that engine oil revenue is influenced by vehicles changing oil, customers and profile interactions by 77.63%.