View Item 
      •   IPB Repository
      • Dissertations and Theses
      • Undergraduate Theses
      • UT - Faculty of Mathematics and Natural Sciences
      • UT - Geophysics and Meteorology
      • View Item
      •   IPB Repository
      • Dissertations and Theses
      • Undergraduate Theses
      • UT - Faculty of Mathematics and Natural Sciences
      • UT - Geophysics and Meteorology
      • View Item
      JavaScript is disabled for your browser. Some features of this site may not work without it.

      Penggunaan Random Forest untuk Estimasi Titik Panas pada Lahan Gambut

      Thumbnail
      View/Open
      Cover (820.9Kb)
      Fulltext (2.500Mb)
      Lampiran (270.3Kb)
      Date
      2024
      Author
      Aulia, Azzahra Zaita Putri
      Taufik, Muh.
      Metadata
      Show full item record
      Abstract
      Titik panas atau hotspot dapat menjadi salah satu indikator tahap awal terjadinya kebakaran hutan dan lahan gambut yang juga menjadi salah satu permasalahan yang cukup masif terjadi di Indonesia. Penentuan hotspot sebagai identifikasi awal kebakaran hutan dan lahan gambut dapat menggunakan teknologi machine learning, salah satunya adalah Random Forest. Penelitian ini bertujuan mengestimasi hotspot menggunakan model Random Forest, serta melihat keakuratan model dan partisi indeks spektral terhadap estimasi hotspot. Penggunaan Random Forest dalam penelitian ini memanfaatkan berbagai indeks spektral sebagai prediktor terhadap estimasi hotspot di wilayah KHG Sungai Kahayan-Sungai Sebangau, Kalimantan Tengah, meliputi NDVI, NDMI, NDWI, EVI, EVI2, NBR, dan NBR2 yang diekstrak dari data citra satelit Sentinel-2A. Model Random Forest memanfaatkan indeks spektral untuk mengestimasi kelas pada setiap lokasi hotspot, meliputi kelas low, nominal, dan high. Model menunjukkan performa yang cukup rendah pada kelas low yang berada di kisaran 30%, sebaliknya model menunjukkan performa yang cukup baik dalam mengestimasi hotspot pada kelas nominal dan high yang berada di kisaran 60-70%. Nilai overall accuracy (OA) juga menunjukkan hasil yang cukup tinggi pada data testing dan validasi, yaitu mencapai 65% dengan indikasi model tidak mengalami overfitting. Tingkat partisi indeks spektral tertinggi adalah NBR2 dan partisi indeks terendah adalah EVI1. Estimasi hotspot dengan random forest juga menunjukkan indikasi adanya bias akibat class imbalance. Penentuan hotspot menggunakan model Random Forest diharapkan dapat berkontribusi terhadap kajian mengenai pemantauan kebakaran hutan dan lahan gambut.
       
      Hotspots can serve as early indicators of forest and peat fires, which are significant issues frequently occurring in Indonesia. Identifying hotspots as an early warning of forest and peat fires can be effectively conducted using machine learning technology, particularly the Random Forest algorithm. This study aims to estimate hotspots using the Random Forest model and assess the accuracy of the model and the spectral index partitions in estimating hotspots. The application of Random Forest in this research leverages various spectral indices as predictors for hotspot estimation in the KHG Sungai Kahayan-Sungai Sebangau area, Central Kalimantan. The indices used include NDVI, NDMI, NDWI, EVI, EVI2, NBR, and NBR2, all extracted from Sentinel-2A satellite imagery data. Random Forest model classifies each hotspot location into low, nominal, and high categories based on spectral indices. The model demonstrates lower performance for the low category, with an accuracy of approximately 30%, but shows better performance for the nominal and high categories, with accuracies ranging from 60% to 70%. The overall accuracy (OA) of the model is relatively high for both testing and validation datasets, reaching 65%, indicating that the model does not suffer from overfitting. The highest spectral index partition was NBR2, while the lowest was EVI1. The estimation of hotspots using the Random Forest model also indicated potential bias due to class imbalance. Utilizing the Random Forest model for hotspot detection is expected to contribute to the study of monitoring and management of forest and peat fires.
       
      URI
      http://repository.ipb.ac.id/handle/123456789/153760
      Collections
      • UT - Geophysics and Meteorology [1720]

      Copyright © 2020 Library of IPB University
      All rights reserved
      Contact Us | Send Feedback
      Indonesia DSpace Group 
      IPB University Scientific Repository
      UIN Syarif Hidayatullah Institutional Repository
      Universitas Jember Digital Repository
        

       

      Browse

      All of IPB RepositoryCollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

      My Account

      Login

      Application

      google store

      Copyright © 2020 Library of IPB University
      All rights reserved
      Contact Us | Send Feedback
      Indonesia DSpace Group 
      IPB University Scientific Repository
      UIN Syarif Hidayatullah Institutional Repository
      Universitas Jember Digital Repository