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

      Prediksi Jumlah Hotspot di Kalimantan Menggunakan Model Long Short-Term Memory Berdasarkan Indikator Iklim

      Thumbnail
      View/Open
      Cover (488.2Kb)
      Fulltext (1.682Mb)
      Lampiran (503.8Kb)
      Date
      2024
      Author
      Fauzan, Muhammad Daryl
      Nurdiati, Sri
      Julianto, Mochamad Tito
      Metadata
      Show full item record
      Abstract
      Forest and land fires have become a concerning phenomenon in Indonesia. As mitigation, hotspots are used as an early stage to identify fires. In an effort to predict the number of hotspots, deep learning models such as Long Short-Term Memory (LSTM) can be used. The purpose of this research is to build the best LSTM model architecture to predict the number of hotspots in Kalimantan based on climate indicators. The model is trained using the glorot uniform method as a weight initiation method, Nadam optimizer with a learning rate of 0.001, and early stopping as model regularization on training data with a proportion of 80%. The results show that the best LSTM model in this research is a model with 2 LSTM layers and 4 dense layers trained to minimize the Root Mean Squared Error (RMSE) as a loss function. The model uses a window of size 12 and a hidden layer with 64 nodes. On testing data, the model is able to predict the number of hotspots with an explained variance score of 83.4%.
       
      Kebakaran hutan dan lahan sudah menjadi fenomena yang memprihatinkan di Indonesia. Sebagai mitigasi, titik panas (hotspot) digunakan sebagai tahap awal untuk mengidentifikasi kebakaran. Dalam upaya memprediksi jumlah hotspot, model deep learning seperti Long Short-Term Memory (LSTM) dapat digunakan. Tujuan penelitian ini adalah membangun arsitektur model LSTM terbaik untuk memprediksi jumlah hotspot di Kalimantan berdasarkan indikator iklim. Model dilatih menggunakan metode glorot uniform sebagai metode inisiasi bobot, optimizer Nadam dengan learning rate 0,001, dan early stopping sebagai regularisasi model pada data training dengan proporsi sebesar 80%. Hasil penelitian menunjukkan bahwa model LSTM yang terbaik pada penelitian ini adalah model dengan 2 layer LSTM dan 4 dense layer yang dilatih untuk meminimumkan Root Mean Squared Error (RMSE) sebagai loss function. Model tersebut menggunakan window berukuran 12 dan hidden layer dengan 64 node. Pada data testing, model mampu memprediksi jumlah hotspot dengan explained variance score sebesar 83,4%.
       
      URI
      http://repository.ipb.ac.id/handle/123456789/154270
      Collections
      • UT - Mathematics [1487]

      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