KLASIFIKASI STATUS RUMAH TANGGA DI PROVINSI BENGKULU MENGGUNAKAN MULTIVARIATE ADAPTIVE REGRESSION SPLINE (MARS)

Abstract

Poverty is a global issue that captures the attention of governments in any country because it is a complex population-related problem. Poverty is a high-dimensional case, involving numerous predictor variables that interact with each other. This study was conducted to obtain a model that is capable of classifying household in Bengkulu Province. Multivariate Adaptive Regression Spline (MARS) is one of the methods used for classification of high-dimensional data. The MARS model is performed with combining Maximum Base Function (BF), Minimal Observation (MO), and Maximum Interaction (MI) with a small Generalized Cross Validation (GCV) by trial and error.  The data used in this study is data from the 2022 National Socioeconomic Survey sourced from the Central Statistics Agency of Bengkulu Province. The variables used is the proverty status of households classified as poor and not poor households as a response variable as well as several predictor variables. The results of this study indicate that it produces a MARS model with a combination of Basis Function (BF) = 48, Maximum Interaction (MI) = 3, and Minimum Observation (MO) = 0 which has the minimum GCV criteria of 0.06799. The results of the accuracy evaluation of the classification obtained an accuracy of 91.65%.