Evaluation of Machine Learning Models for Mapping Food Crops using Sentinel-2A Imagery in West Java, Indonesia

¹ Mapping Survey and Geographic Information Study Program, Faculty of Social Sciences Education, Universitas Pendidikan Indonesia, Jl. Dr. Setiabudhi No. 229, Bandung, Indonesia
² Doctoral Program in Geography, Faculty of Geography, Universitas Gadjah Mada, Yogyakarta, Indonesia
³ Departement of Geography Information Science, Faculty of Geography, Universitas Gadjah Mada, Bulaksumur, Yogyakarta 55281, Indonesia
⁴ Geography Information Science Study Program, Faculty of Social Sciences Education, Universitas Pendidikan Indonesia, Jawa Barat, Indonesia.

Abstract

Data on the distribution patterns and locations of food crops are crucial for monitoring and controlling the sustainability of agricultural resources and guaranteeing food security. Plant classification based on machine learning has been widely used to detect food crop areas. However, there are still challenges in mapping plant types and plant area effectively and efficiently. The aim of this research is to evaluate machine learning models in mapping and calculating the area of food crops (rice) in West Java Province, Indonesia. Google Earth Engine is used in this study as a big data cloud computing platform for remote sensing. Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) Sentinel2A imagery is utilized to employ time series data as input characteristics for the three most popular machine learning models: Support Vector Machine (SVM), Random Forest (RF), and Classification and Regression Trees (CART). The research results show that the three machine learning models are able to map and calculate the area of food crops in West Java, Indonesia. The RF algorithm produces the highest overall accuracy rate (98.51%) and is the fastest in the accuracy assessment and image classification process compared to the SVM and CART algorithms.