Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/161249
Title: Algoritma Pembelajaran Mesin Pendeteksi Wana-Tani di Kecamatan Lore Barat, Kabupaten Poso, Provinsi Sulawesi Tengah
Other Titles: Machine Learning Algorithm for Agroforestry Detection in West Lore District, Poso Regency, Central Sulawesi Province
Authors: Jaya, I Nengah Surati
Diefda, Genthamaury
Issue Date: 2025
Publisher: IPB University
Abstract: Penelitian ini menjelaskan tentang pembangunan algoritma pohon keputusan (decision tree) pembelajaran mesin pendeteksian wana-tani untuk mengidentifikasi perhutanan sosial yang terdapat di Kecamatan Lore Barat. Pembangunan model algoritma dan deteksi wana-tani dilakukan dengan menggunakan citra SPOT 7 dalam menduga tutupan lahan dan membangun data tambahan melalui peubah indeks vegetasi (NDVI, NRGI, VARI) serta data geososio biofisik (slope, elevasi, pemukiman, jalan, sungai). Hasil dari penelitian ini berupa model algoritma terbaik dalam melakukan klasifikasi tutupan lahan untuk mendeteksi wana-tani dengan parameter terbaik berupa information gain. Parameter ini menghasilkan model algoritma terbaik tanpa pre-prunning dan tanpa prunning dengan maximal depth sebesar 10, minimal leaf size sebesar 60, pre-prunning alternative sebesar 20, dan minimal size for split sebesar 41. Kajian ini menghasilkan overall accuracy sebesar 92% dan kappa accuracy sebesar 90% dengan prediksi kelas agroforestri sebesar 77.2% untuk user accuracy dan 75.5% untuk kappa accuracy.
This research describes the development of a decision tree algorithm for machine learning wana-tani detection to identify social forestry in West Lore District. The development of algorithm model and wana-tani detection were conducted using SPOT 7 imagery in estimating land cover and using additional data through vegetation index variables (NDVI, NRGI, VARI) and geosocio-biophysical data (slope, elevation, settlement, road, river). The result of this research provided the best algorithm model in performing land cover classification to detect wana-tani with the best parameter in the form of information gain. This parameter produced the best algorithm model without pre-prunning and without prunning with a maximum depth of 10, minimum leaf size of 60, pre-prunning alternative of 20, and minimum size for split of 41. This research produced an overall accuracy of 92% and kappa accuracy of 90% with a prediction of agroforestry class of 77.2% for user accuracy and 75.5% for kappa accuracy.
URI: http://repository.ipb.ac.id/handle/123456789/161249
Appears in Collections:UT - Forest Management

Files in This Item:
File Description SizeFormat 
cover_E1401201066_fa864dab52734a339d1afd77d1c66d8f.pdfCover366.67 kBAdobe PDFView/Open
fulltext_E1401201066_9a08b38e6f9f45b29d044ce00aa82347.pdf
  Restricted Access
Fulltext2.64 MBAdobe PDFView/Open
lampiran_E1401201066_dac73aeb604442f38ddd684a8f509721.pdf
  Restricted Access
Lampiran286.77 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.