Dinamika Perubahan Kesehatan Mangrove Di Pantai Timur Provinsi Jambi Berbasis Machine Learning Decision Tree

Abstract

Penelitian ini mendeskripsikan pengembangan algoritma pembelajaran mesin berbasis pohon keputusan untuk mendeteksi tingkat kesehatan mangrove dan menganalisis dinamika perubahannya dari tahun 2019 hingga 2025, sebagai respons terhadap tekanan antropogenik seperti pemanfaatan kayu dan konversi lahan di Pantai Timur Provinsi Jambi yang telah memicu degradasi vegetasi secara signifikan. Analisis data spasial dilakukan menggunakan citra satelit Sentinel-2A multi-temporal yang diintegrasikan dengan variabel sosio-geo-biofisik dengan akar pohon keputusan Normalized Difference Water Index (NDWI). Klasifikasi tingkat kesehatan mengacu pada nilai Mangrove Health Index (MHI). Hasil uji akurasi model menunjukkan performa yang optimal, dengan tingkat Akurasi Keseluruhan (Overall Accuracy) mencapai 96,6% dan Akurasi Kappa (Kappa Accuracy) sebesar 96% dengan Cross Validation 5 fold. Sepanjang periode analisis 2019–2025, dinamika perubahan spasial menunjukkan adanya degradasi dan pertumbuhan mangrove.

This research describes the development of a decision tree-based machine learning algorithm to detect mangrove health levels and analyze its change dynamics from 2019 to 2025, in response to anthropogenic pressures such as wood utilization and land conversion on the East Coast of Jambi Province, which have triggered significant vegetation degradation. Spatial data analysis was conducted using multitemporal Sentinel-2A satellite imagery integrated with socio-geo-biophysical variables, with the Normalized Difference Water Index (NDWI) serving as the root node of the decision tree. The health level classification refers to the Mangrove Health Index (MHI) values. The model accuracy test results demonstrated optimal performance, achieving an Overall Accuracy of 96.6% and a Kappa Accuracy of 96% using 5-fold Cross-Validation. Throughout the 2019–2025 analysis period, the spatial change dynamics indicated both the degradation and regrowth of the mangrove ecosystem monitoring.