Algoritma Estimasi Indeks Kesehatan Mangrove dengan Pendekatan Machine Learning di Kabupaten Parigi Moutong, Sulawesi Tengah

Abstract

Penelitian ini menjelaskan pengembangan algoritma machine learning untuk estimasi indeks kesehatan mangrove menggunakan pendekatan decision tree berbasis citra Sentinel-2A di Kabupaten Parigi Moutong, Sulawesi Tengah, Indonesia. Penelitian ini bertujuan mengembangkan algoritma machine learning berbasis decision tree serta mengidentifikasi peubah-peubah utama dalam penilaian Mangrove Health Index (MHI). Algoritma dikembangkan menggunakan indeks spektral yang meliputi Combined Mangrove Recognition Index (CMRI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Intertidal Mangrove Index (NIMI), Normalized Red-Green Vegetation Index (NRGI), Normalized Difference Moisture Index (NDMI), dan Normalized Ratio Built-up Index (NRBI), yang diintegrasikan dengan peubah sosio-geo-biofisik berupa elevasi, jarak dari jalan, sungai, garis pantai, dan substrat. Penelitian ini mengungkapkan model decision tree terbaik menghasilkan Overall Accuracy sebesar 93,7% dan Kappa Accuracy sebesar 0,93, dengan precision sebesar 83,6–95,7% dan recall sebesar 88,0–92,2% pada kelas Indeks Kesehatan Mangrove.

This paper describes the development of A Machine Learning-Based Algorithm for Estimating the Mangrove Health Index in Parigi Moutong Regency, Central Sulawesi, Indonesia. The objective of the study was to develop a decision tree machine learning algorithm for identifying the key variables on assessing the Mangrove Health Index (MHI). The algorithm was developed using spectral indices, including the Combined Mangrove Recognition Index (CMRI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Intertidal Mangrove Index (NIMI), Normalized Red-Green Vegetation Index (NRGI), Normalized Difference Moisture Index (NDMI), and Normalized Ratio Built-up Index (NRBI), integrated with socio-geo-biophysical variables consisting of elevation, distance from roads, rivers, coastline, and substrate. The study revealed that the optimal decision tree model achieved an Overall Accuracy of 93.7% and a Kappa Accuracy of 0.93. The precision for the MHI is ranging between 83.6.% to 95.7%, while the recall are between 88% to 92.2%.