Prediksi Curah Hujan di Lombok Menggunakan Ekstraksi Fitur dan Model PSO-BiLSTM

Dewi, Arin Rosana (2026) Prediksi Curah Hujan di Lombok Menggunakan Ekstraksi Fitur dan Model PSO-BiLSTM. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221010-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (12MB) | Request a copy

Abstract

Fenomena perubahan iklim global telah menyebabkan frekuensi dan intensitas cuaca ekstrem meningkat, khususnya curah hujan di Lombok karena karakteristik geografis serta pengaruh perubahan iklim global. Kondisi ini meningkatkan risiko bencana hidrometeorologi yang berdampak besar terhadap masyarakat dan infrastruktur. Prediksi curah hujan yang akurat diperlukan untuk mitigasi bencana, namun metode konvensional masih memiliki keterbatasan dalam menangkap hubungan data meteorologi yang nonlinier, nonstasioner, dan terdapat noise. Penelitian ini mengembangkan model untuk memprediksi curah hujan dengan menggunakan penanganan ekstraksi fitur berbasis VMD-KPCA dan model PSO-BiLSTM pada data cuaca multiatribut di Lombok. Data mencakup indeks iklim lokal (curah hujan, temperatur, kelembaban, lama penyinaran matahari, kecepatan angin) dan indeks iklim global (SOI, NINO, IPWP, DMI) dalam rentang waktu 20 tahun. Variational Mode Decomposition (VMD) digunakan untuk mendekomposisi 14 fitur masukan menjadi 70 Intrinsic Mode Functions (IMF) dengan K = 5, kemudian direduksi secara nonlinier menggunakan Kernel Principal Component Analysis (KPCA) menjadi 18 komponen utama. Particle Swarm Optimization (PSO) mengoptimasi hyperparameter Bidirectional Long Short-Term Memory (BiLSTM) dengan fungsi fitness Huber Loss, sehingga diperoleh konfigurasi hidden unit sebesar 256, hidden layer sebesar 1, dropout rate sebesar 0,001, learning rate sebesar 0,001, regularisasi L2 sebesar 10−5, dan batch size sebesar 25. Hasil prediksi curah hujan pada data uji menunjukkan R2 sebesar 0,9825, MAE sebesar 0,4178 mm, RMSE sebesar 0,6541 mm, dan MAPE sebesar 20,26%.
========================================================================================================================================
Global climate change has increased the frequency and intensity of extreme weather events, particularly rainfall in Lombok due to its geographical characteristics and the influence of global climate change. This condition increases the risk of hydrometeorological disasters that have major impacts on communities and infrastructure. Accurate rainfall prediction is required for disaster mitigation however, conventional methods still have limitations in capturing nonlinear, nonstationary, and noisy meteorological data. This study develops a rainfall prediction model using VMD-KPCA-based feature extraction and a PSO-BiLSTM model on multiattribute weather data in Lombok. The data consist of local climate indices (rainfall, temperature, humidity, sunshine duration, and wind speed) and global climate indices (SOI, NINO, IPWP, and DMI) over a period of 20 years. Variational Mode Decomposition (VMD) was used to decompose 14 input features into 70 Intrinsic Mode Functions (IMFs) with K = 5, which were then nonlinearly reduced using Kernel Principal Component Analysis (KPCA) into 18 principal components. Particle Swarm Optimization (PSO) optimized the hyperparameters of the Bidirectional Long Short Term Memory (BiLSTM) model using the Huber Loss fitness function, resulting in an optimal configuration of 256 hidden units, 1 hidden layer, dropout rate of 0,001, learning rate of 0,001, L2 regularization of 10−5, and batch size of 25. The rainfall prediction results on the test data achieved R2 of 0,9825, MAE of 0,4178 mm, RMSE of 0,6541 mm, and MAPE of 20,26%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Prediksi Curah Hujan, Ekstraksi Fitur, VMD, KPCA, PSO, BiLSTM ============================================================================================================================================= Rainfall Prediction, Feature Extraction, VMD, KPCA, PSO, BiLSTM
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Arin Rosana Dewi
Date Deposited:	30 Jul 2026 02:47
Last Modified:	30 Jul 2026 02:47
URI:	http://repository.its.ac.id/id/eprint/140959

Actions (login required)

View Item