Electronic Theses and Dissertation

Prediksi curah hujan yang akurat memiliki peran penting dalam mendukung mitigasi bencana hidrometeorologis dan perencanaan sektor pertanian, khususnya di wilayah tropis seperti aceh besar yang memiliki dinamika iklim kompleks. penelitian ini bertujuan untuk membangun model prediksi curah hujan harian menggunakan metode artificial neural network yang dikombinasikan dengan principal component analysis. data yang digunakan merupakan data iklim harian periode juni 2019 hingga juni 2025 yang diperoleh dari badan meteorologi, klimatologi, dan geofisika, mencakup delapan variabel klimatologis yaitu temperatur minimum, temperatur maksimum, temperatur rata-rata, kelembapan udara, lama penyinaranmatahari, kecepatan angin maksimum, kecepatan angin rata-rata, dan arah angin maksimum. pra-pemrosesan data dilakukan melalui teknik normalisasi dan reduksi dimensi menggunakan principal component analysis untuk menyederhanakan struktur input. model artificial neural network dibangun menggunakan arsitektur feedforward dan algoritma backpropagation, dengan variasi jumlah neuron tersembunyi dan laju pembelajaran. hasil evaluasi menunjukkan bahwa model terbaik diperoleh pada konfigurasi tiga neuron tersembunyi dan laju pembelajaran sebesar nol koma satu, dengan nilai mean squared error sebesar 0,0102. model yang menggunakan pca menunjukkan penurunan galat prediksi dibandingkan model ann tanpa pca, sehingga pca terbukti meningkatkan efisiensi pelatihan dan akurasi model. hasil penelitian ini menunjukkan bahwa kombinasi artificial neural network dan principal component analysis dapat menjadi pendekatan yang efektif dalam sistem prediksi curah hujan harian di wilayah aceh besar. kata kunci: curah hujan, artificial neural network, principal component analysis, backpropagation, prediksi iklim

Accurate rainfall prediction plays a crucial role in supporting hydrometeorological disaster mitigation and agricultural planning, especially in tropical regions such as Aceh Besar which experience complex climatic dynamics. This study aims to develop a daily rainfall prediction model using the Artificial Neural Network method combined with Principal Component Analysis. The data used in this research consist of daily climate records from June 2019 to June 2025, obtained from the Meteorology, Climatology, and Geophysics Agency, covering eight climatological variables including minimum temperature, maximum temperature, average temperature, relative humidity, sunshine duration, maximum wind speed, average wind speed, and wind direction at maximum speed. Data preprocessing was carried out through normalization and dimensionality reduction using Principal Component Analysis to simplify the model’s input structure. The Artificial Neural Network model was constructed using a feedforward architecture and trained with the backpropagation algorithm, varying the number of hidden neurons and learning rates. The evaluation results show that the best model configuration uses three hidden neurons and a learning rate of 0.1, yielding the lowest Mean Squared Error value of 0.0102. The model that incorporated Principal Component Analysis produced lower prediction errors compared to the model without PCA, demonstrating improved training efficiency and prediction accuracy. These findings indicate that the integration of Artificial Neural Network and Principal Component Analysis is an effective approach for developing a rainfall prediction system in the Aceh Besar region. Keywords: Rainfall, Artificial Neural Network, Principal Component Analysis, Backpropagation, Climate Prediction