Comparison of main characteristics of food insecurity using classification Tree and Random Forest

Abstract

Since the emerging of big data era, the information and data are grown rapidly. It requires us to have ability to extract the knowledge and information that consisted in this explosion of the data. One of way that can be used for this purpose is by using machine learning method. One of purpose of machine learning implementation is to conduct classification analysis and to identify variable importance that contribute in the research. It’s conducted the comparative study between two machine learning classification methods named classification tree and random forest method. This study is implemented on Indonesian Socioeconomic Survey (SUSENAS) 2020 in Aceh Province. The purpose of the study is to identify the optimum method between both and to identify the characteristics of food insecure household. The optimum method obtained by comparing the AUC value. The results obtained is random forest outperformed classification tree with the AUC value of random forest method is 0,718 and classification tree method is 0,668. The rank of variable importance of the optimum method is the type of cooking fuel used in the household, the area of house floor, education level of head of household, number of savers in a household, and the type of house floor.