Electronic Theses and Dissertation

Perubahan iklim berdampak signifikan terhadap kondisi atmosfer, termasuk meningkatnya frekuensi kekeringan yang memengaruhi ketersediaan air dan ketahanan pangan. Kota Banda Aceh mengalami variabilitas iklim yang tinggi, sehingga penelitian ini bertujuan untuk menentukan metode imputasi terbaik dalam menangani data iklim yang hilang serta memilih fungsi kernel optimal pada Support Vector Regression (SVR) dalam memprediksi indeks kekeringan. Penelitian ini menggunakan data iklim harian Kota Banda Aceh periode 2010–2024 yang bersumber dari website Badan Meteorologi, Klimatologi, dan Geofisika (BMKG). Data ini mencakup variabel temperatur, kelembapan, curah hujan, lamanya penyinaran matahari, dan kecepatan angin. Metode single imputation yang diuji mencakup Last Observation Carried Forward (LOCF), Next Observation Carried Backward (NOCB), dan Baseline Observation Carried Forward (BOCF). Evaluasi kinerja metode imputasi dilakukan menggunakan metrik Mean Absolute Error (MAE), Mean Absolute Scaled Error (MASE), Relative Absolute Error (RAE), dan Weighted Mean Absolute Percentage Error (WMAPE). Setelah imputasi menggunakan metode terbaik, diterapkan SVR dengan kernel linier, polinomial, Radial Basis Function (RBF), dan sigmoid. Evaluasi kinerja fungsi kernel dilakukan menggunakan metrik Root Mean Square Error (RMSE), Median Absolute Error (MedAE), dan MAE. Hasil penelitian menunjukkan bahwa metode NOCB memberikan performa terbaik dalam menangani data hilang pada data iklim Kota Banda Aceh dengan kesalahan terendah di mana nilai MAE, MASE, RAE, dan WMAPE secara berturut turut sebesar 0,965; 0,023; 0,040; dan 4,047. Sementara itu, model SVR dengan kernel linier memiliki nilai kesalahan paling rendah berdasarkan metrik RMSE, MedAE, dan MAE secara berturut-berturut 0,300; 0,216; dan 0,248 untuk SPI serta 0,256; 0,186; dan 0,208 untuk SPEI yang menunjukkan bahwa kernel linier paling optimal dalam memprediksi indeks kekeringan di Kota Banda Aceh.

Climate change has a significant impact on atmospheric conditions, including increased frequency of droughts that affect water availability and food security. Banda Aceh City experiences high climate variability, so this study aims to determine the best imputation method in handling missing climate data and selecting the optimal kernel function in Support Vector Regression (SVR) in predicting the drought index. This study uses daily climate data for Banda Aceh City for the period 2010-2024 sourced from the website of the Meteorology, Climatology and Geophysics Agency (BMKG). This data includes variables of temperature, humidity, rainfall, sunshine duration, and wind speed. The single imputation methods tested include Last Observation Carried Forward (LOCF), Next Observation Carried Backward (NOCB), and Baseline Observation Carried Forward (BOCF). Performance evaluation of the imputation methods was conducted using the Mean Absolute Error (MAE), Mean Absolute Scaled Error (MASE), Relative Absolute Error (RAE), and Weighted Mean Absolute Percentage Error (WMAPE) metrics. After imputation using the best method, SVR is applied with linear, polynomial, Radial Basis Function (RBF), and sigmoid kernels. Evaluation of kernel function performance is carried out using the Mean Square Error (RMSE), Median Absolute Error (MedAE), and MAE metrics. The results show that the NOCB method provides the best performance in handling missing data in Banda Aceh City climate data with the lowest error where the MAE, MASE, RAE, and WMAPE values are 0.965, 0.023, 0.040, and 4.047, respectively. Meanwhile, the SVR model with a linear kernel has the lowest error value based on the RMSE, MedAE, and MAE metrics of 0.300, 0.216, and 0.248 for SPI and 0.256, 0.186, and 0.208 for SPEI, respectively, indicating that the linear kernel is the most optimal in predicting the drought index in Banda Aceh City.