Feature Selection pada Indikator Indeks Ekonomi Hijau di Indonesia dengan Machine Learning

Abstract

Green economy policies are crucial for all countries to ensure that economic activities progress while preserving environmental sustainability. The success of such policies is measured by the Green Economy Index, which in 2020 recorded a national score of 59.17 with 15 indicators, while provincial-level indicators are still being developed. This study analyzes 18 provincial indicators to identify the main factors influencing the Green Economy Index using LASSO regression. This method was chosen for its ability to efficiently perform feature selection, address multicollinearity, and reduce overfitting risks. The dataset includes 18 indicators and index values from 34 provinces. The results show that 15 indicators significantly affect the index. The developed model demonstrates good performance with an RMSE of 1.23 for the training set and 2.29 for the testing set. The R² values of 95.6% (training) and 85.98% (testing) indicate strong predictive capability. Moreover, surface water quality is identified as the most influential indicator. These findings are expected to support data-driven policymaking in strengthening the green economy at the provincial level.