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 dengan Metode Regresi Proses Gaussian Berdasarkan Indikator Iklim

      Thumbnail
      View/Open
      Cover (699.5Kb)
      Fulltext (2.416Mb)
      Lampiran (279.7Kb)
      Date
      2024
      Author
      Nurdianto, Hari
      Nurdiati, Sri
      Julianto, Mochamad Tito
      Metadata
      Show full item record
      Abstract
      Hotspot merupakan suatu indikator kebakaran hutan yang sangat dipengaruhi oleh indikator iklim lokal dan global. Kebakaran hutan merupakan salah satu bencana alam yang cukup rutin terjadi di Indonesia terutama di pulau Kalimantan. Penelitian ini mengembangkan model dengan low hyperparameter complexity untuk memprediksi jumlah hotspot berdasarkan indikator iklim. Tujuan penelitian ini yaitu mengonstruksi model regresi proses Gaussian menggunakan data jumlah hotspot di Kalimantan dengan fungsi kernel ARD eksponensial kuadrat dan melakukan tuning pada hyperparameter yakni nilai varian dari noise pada fungsi kernel. Tuning hyperparameter dilakukan dengan tiga metode optimisasi yaitu Bayesian optimization, grid search, dan random search. Hasilnya, model regresi proses Gaussian terbaik diperoleh menggunakan metode Bayesian optimization dan random search dengan nilai varian dari noise masing – masing sebesar 1401.3 dan 1128.2. Metode prediksi regresi proses Gaussian yang telah dioptimisasi menggunakan dua metode tersebut menghasilkan nilai metrik akurasi pada data testing yaitu nilai RMSE, MAE, dan R-squared masing-masing sebesar 844.47, 354.29, 54.52% dan 846.58, 350.93, 54,29%.
       
      An hotspot is an indicator of forest fires heavily influenced by the local and global climate indicators. Forest fires are a natural disaster that regularly occurs in Indonesia, especially on the island of Kalimantan. This study develops a model with low hyperparameter complexity to predict the number of hotspots based on climate indicators. The purpose of this research is to construct a model using Gaussian Process Regression (GPR) with data on hotspot numbers in Kalimantan employing the ARD squared exponential kernel functions and tuning hyperparameters such as the variance value of noises in the kernel function. The hyperparameter tuning was performed using three optimizations. The methods are the Bayesian optimization, the grid search, and the random search. The results shows that the best GPR model was obtained using the Bayesian optimization and the random search, with sigma values of 1401.3 and 1128.2, respectively. The optimized GPR prediction methods using these two methods produce in accuracy metrics of RMSE, MAE, and Rsquared values of 844.47, 354.29, 54.52% and 846.58, 350.93, 54.29% respectively.
       
      URI
      http://repository.ipb.ac.id/handle/123456789/154715
      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