COMPARATIVE ACCURACIES USING MACHINE LEARNING MODELS FOR MAPPING OF SUGARCANE PLANTATION BASED ON SENTINEL-2A IMAGERY IN KEDIRI AREA, EAST JAVA

Aug 6, 2026

Data collection in smallholder sugarcane plantations remains highly sensitive to the subjectivity of informants and data collectors, while company-level surveys suffer from low response rates. These issues can degrade estimate precision for sugarcane plantation area. This study recognizes sugarcane fields in Kediri Regency and Kediri Municipality, East Java, using Sentinel-2A imagery with multiple machine learning models. It compares LightGBM against CART, SVM, Random Forest, and XGBoost, using hyperparameter tuning with random search and stratified 10-fold cross validation. Labeling relies on Google Street View imagery, and predictors include NDVI, NDWI, NDBI, EVI, and elevation; LightGBM achieves 98% accuracy with Cohen’s kappa 97.7%. Estimated areas for September 2022 are 18,897.6 ha and 571.87 ha.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Abstract

Introduction

Data sensitivity and response-rate limitations

Need for up-to-date plantation data

Remote sensing and machine learning background

Methods (from abstract)

Machine learning models and comparison

Feature labeling and predictor variables

Hyperparameter tuning and cross validation

COMPARATIVE ACCURACIES USING MACHINE LEARNING MODELS FOR MAPPING OF SUGARCANE PLANTATION BASED ON SENTINEL-2A IMAGERY IN KEDIRI AREA, EAST JAVA Ridson Al Farizal Pulungan1, Rani Nooraeni2 12Departemen of Statistic, Politeknik Statistika STIS, Jakarta, Indonesia e-mail: [ridsonalfarizal15@gmail.com](ridsonalfarizal15@gmail.com) Received: 18-01-2023; Revised: 25-12-2023; Approved: 26-01-2024 Abstract. Data collection in smallholder sugarcane plantations is still very sensitive to the subjectivity of informants and data collectors. In the meantime, the problem with data collection on sugarcane plantation companies is a low response rate. This situation can reduce the precision of the estimates that are produced. Consequently, the goal of this research is to recognize sugarcane fields using the machine learning models on Sentinel-2A satellite imagery in Kediri Area that covering Kediri Regency and Kediri Municipality, East Java. Along with developing machine learning algorithms, this research will evaluate how well LightGBM performs when compared to other algorithms, including CART, SVM, Random Forest, and XGBoost. Each model employed hyperparameter tuning with random search and stratified 10-fold cross validation to avoid overfitting. The process of labelling satellite imagery using images from Google Street View, then predictor variables used are NDVI, NDWI, NDBI, EVI, and elevation. The most accurate classification model obtained was LightGBM, with a 98% accuracy and a cohen’s kappa of 97.7% . The estimated area of sugarcane plantations in the Kediri Regency and Kediri Municipality in September 2022 is 18,897.6 ha and 571.87 ha. Keywords: remote sensing, CART, SVM RBF kernel, SVM polynomial kernel, Random Forest, XGBoost, LightGBM 1 INTRODUCTION Sugarcane (Saccharum officinarum) is a member of the Gramineae family, which includes grasses. The sugar and monosodium glutamate (MSG) industries utilize the water extracted from sugarcane stalks as a raw ingredient. (Syathori & Verona, 2020) . Thousands of factory workers and sugarcane farmers depend on the sugarcane and MSG industries for a living. Furthermore, sugar has become a necessity for most Indonesians (Sulaiman et al. , 2018) . Indonesia's annual sugar consumption continues to rise (BPS, 2022) . However, the sugarcane plantations area did not rise considerably; in fact, it decreased because of land conversion. As part ofits attempts to establish national food security, the Indonesian government is attempting to attain self-sufficiency in sugar production (Sulaiman et al. , 2018) . Accurate and up to date data on sugarcane plantations is required to progress sugarcane plantations in Indonesia. Statistical Agency (BPS) and the Directorate General of Plantations are the two primary data sources for sugarcane plantations in Indonesia. BPS obtained data related to sugarcane plantation companies using Computer Assisted Web Interviewing (CAWI)-based self-enumeration and by interviewing companies that had not filled out the form. (BPS, 2022) . Meanwhile, the Directorate General of Plantations collects statistics on smallholder plantations through field officers' estimations. Officers will collect data on planting methods, population density per hectare, land area (distinct from planting area), and other factors. The sources of information included planters, farmer groups, village officials, and others. Officers will estimate the area based on this data in accordance with the Guidelines for Implementing Plantation Commodity Data Management (PDKP) (Kementrian Pertanian, 2013) . However, until recently, data gathering on smallholder plantations was very sensitive to informant and data collector subjectivity (Ruslan & Prasetyo, 2021) . The guidelines for creating predictions that are not up to date can provide estimates that are not in agreement with the present circumstance. Moreover, the problem with data gathering on sugarcane plantation companies is a l

FAQs

What is the best-performing model and the estimated sugarcane area?

LightGBM is the most accurate, reaching 98% accuracy with Cohen’s kappa 97.7%. The estimated September 2022 areas are 18,897.6 ha in Kediri Regency and 571.87 ha in Kediri Municipality.

Aug 6, 2026

Upload by Lucas Martin