WIJAYANTO, Ahmad and Sugiharto, Aris and Santoso, Rukun (2024) PREDIKSI CURAH HUJAN BERPOTENSI BANJIR MENGGUNAKAN ALGORITMA LONG SHORT-TERM MEMORY (LSTM) DAN ISOLATION FOREST. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text COVER.pdf Download (1MB)
	Text BAB I.pdf Download (451kB)
	Text BAB II.pdf Download (943kB)
	Text BAB III.pdf Restricted to Repository staff only Download (784kB)
	Text BAB IV.pdf Restricted to Repository staff only Download (4MB)
	Text BAB V.pdf Restricted to Repository staff only Download (433kB)
	Text Daftar Pustaka.pdf Download (370kB)
	Text LAMPIRAN.pdf Restricted to Repository staff only Download (856kB)

Abstract

Permasalahan banjir yang sering terjadi di berbagai daerah dapat disebabkan oleh faktor curah hujan yang sulit diprediksi. Penelitian ini bertujuan untuk menghasilkan model prediksi dan peramalan jangka panjang curah hujan menggunakan algoritma Long Short Term Memory (LSTM) serta identifikasi pengambilan keputusan curah hujan berpotensi banjir menggunakan Isolation Forest. Algoritma LSTM digunakan untuk memprediksi curah hujan dengan hasil training dan validation Mean Squared Error (MSE) sebesar 17,04 dan Root Mean Squared Error (RMSE) sebesar 4,12. Probabilitas curah hujan dalam jangka panjang diperoleh dengan penggunaan Gaussian Noise pada LSTM yang dipengaruhi oleh tingkat akurasi terbaik. Selanjutnya, algoritma Isolation Forest digunakan untuk mengidentifikasi curah hujan yang berpotensi menyebabkan banjir. Hasil dari prediksi ini dapat digunakan untuk menentukan pengambilan keputusan mengenai kejadian curah hujan berpotensi banjir. Dengan hasil probabilitas curah hujan dalam jangka panjang dan identifikasi dini potensi banjir, dapat dimanfaatkan untuk merancang infrastruktur yang lebih tahan banjir, seperti sistem drainase yang lebih baik, pembangunan fasilitas pengendalian banjir yang sesuai, serta perhitungan kekuatan struktur di daerah rawan banjir.
Kata kunci : prediksi, LSTM, Isolation Forest

The recurring flood problems in various regions can be attributed to the unpredictable nature of rainfall patterns. This research aims to develop a long-term rainfall prediction and forecasting model using the Long Short-Term Memory (LSTM) algorithm, as well as identify rainfall decisions with potential flood risks using the Isolation Forest algorithm. The LSTM algorithm was employed to predict rainfall, with the training and validation results yielding a Mean Squared Error (MSE) of 17.04 and a Root Mean Squared Error (RMSE) of 4.12. The long-term rainfall probability was obtained by incorporating Gaussian Noise into the LSTM, which was influenced by the best accuracy level. Furthermore, the Isolation Forest algorithm was used to identify rainfall patterns with the potential to cause floods. The results of this prediction can be used to determine decision-making regarding rainfall events with flood potential. By obtaining long-term rainfall probability and early identification of potential flood risks, the information can be utilized to design infrastructure that is more resilient to floods, such as improved drainage systems, the construction of appropriate flood control facilities, and the consideration of structural strength in flood-prone areas.
Keywords : prediksi, LSTM, Isolation Fores

Actions (login required)

View Item