Penerapan Metode CART pada Pengklasifikasian Bekerja dan Pengangguran di Kabupaten Subang

Abstract

Unemployment is a complex problem faced by developing countries,
including Indonesia. The high unemployment rate in Indonesia impacts poverty, so
that the government seeks to carry out economic development. Subang is one of the
districts that contributed 8,68 percent of the open unemployment rate in 2019 and
increased by 9,48 percent in 2020. The incessant growth of industrial estates and
smart city program development in Subang is one of the efforts to reduce
unemployment. This study used a classification and regression tree (CART) to
determine the factors that influenced unemployment status in Subang Regency. The
advantage of the CART method is easy to interpret the results of the analysis.
However, the accuracy of the classification tree is relatively low due to data
imbalance. Therefore, this study used SMOTE method to deal with this problem.
The optimal classification tree was formed from 17 terminal nodes and 6
explanatory variables. 7 terminal nodes represent work as work, and 10 terminal
nodes represent unemployment as unemployment. The 6 explanatory variables
consist of marital status (X3), attending job training (X5), the position in the family
(X4), the education level (X2), gender (X1), and age (X6).