Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/132738
Title: Pendugaan Produktivitas Padi Menggunakan Citra Satelit Sentinel-2 dengan Pendekatan Machine Learning Berbasis Indeks Vegetasi
Other Titles: Estimating Rice Productivity Using Sentinel-2 Satellite Imagery with a Machine Learning Approach Based on Vegetation Index
Authors: Seminar, Kudang Boro
Supriyanto, Supriyanto
Adyatma, Naufal Dian
Issue Date: 2023
Publisher: IPB University
Abstract: Informasi ketersediaan beras di masa mendatang mendukung kebijakan pemerintah guna mewujudkan ketahanan pangan yang berkelanjutan. Salah satu metode yang digunakan untuk mendukung ketersediaan beras di masa mendatang yaitu penggunaan teknologi penginderaan jauh. Penelitian ini bertujuan untuk menduga produktivitas padi berdasarkan nilai normalized difference vegetation index (NDVI) citra satelit Sentinel-2 melalui pendekatan machine learning decision tree. Data produksi padi, varietas, tanggal tanam, tanggal panen, dan koordinat sawah didapat dengan melakukan wawancara kepada petani serta survei lapang. Koordinat sawah kemudian dikonversi menjadi luas area dan dibuat produktivitas padi per petakan sawah berdasarkan data yang sudah didapat. Nilai NDVI per petakan sawah selama satu musim tanam kemudian dihubungkan dengan produktivitas padi melalui machine learning decision tree. Model terbaik pada pendugaan gabungan varietas Inpari 32 dan MR 219 memiliki nilai MAPE 5,077% dengan akurasi 94,923%; RMSE ± 0,391 ton/ha; dan nilai R2 0,776. Model pendugaan terbaik pada varietas Inpari 32 memiliki nilai MAPE 4,16% dengan akurasi 95,84%; RMSE ± 0,262 ton/ha; dan R2 0,894. Model pendugaan terbaik pada varietas MR 219 pada varietas MR 219 memiliki nilai MAPE 1,339% dengan akurasi 98,661%; RMSE sebesar ± 0,103 ton/ha; dan R2 0,982.
Information on future rice availability supports government policies for sustainable food security. One method used to support rice availability in the future is the use of remote sensing technology. This study aims to predict the productivity of rice based on normalized difference vegetation index (NDVI) of the Sentinel-2 satellite image through a machine learning decision tree approach. Rice production, plants variety, planting date, harvest date, and field coordinates were obtained by conducting interviews with farmers as well as field surveys. Paddy field coordinates then converted into area, and productivity per square area is generated based on the obtained data. The NDVI value per pitch for a single growing season is linked to the productivity of paddy through the machine learning decision tree. The best model on the combined varieties of Inpari 32 and MR 219 has a MAPE value of 5,077% with an accuracy 94,923%; RMSE ± 0,391 tons/ha; and R² 0,776. The best model on Inpari 32 varieties has a MAPE value of 4,16% with an accuracy 95,84%; RMSE ± 0,262 tons/ha; and R2 0,894. The best model on MR 219 varieties has a MAPE value of 1,339% with an accuracy 98,661%; RMSE ± 0,103 tons/ha; and R2 0,982.
URI: http://repository.ipb.ac.id/handle/123456789/132738
Appears in Collections:UT - Agricultural and Biosystem Engineering

Files in This Item:
File Description SizeFormat 
Cover, Lembar Pengesahan, Prakata, Daftar isi.pdf
  Restricted Access
Cover932.53 kBAdobe PDFView/Open
F14190097_Naufal Dian Adyatma.pdf
  Restricted Access
Fullteks3.94 MBAdobe PDFView/Open
Lampiran.pdf
  Restricted Access
Lampiran2.37 MBAdobe PDFView/Open


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