Prediks Curah Hujan Harian Menggunakan Artificial Neural Networ dengan BOHB di Sepinggan Balikpapan

Puspitasari, Dilla (2026) Prediks Curah Hujan Harian Menggunakan Artificial Neural Networ dengan BOHB di Sepinggan Balikpapan. Bachelor thesis, Institut Teknologi Kalimantan.

	Text 11221031_cover.pdf Restricted to Repository staff only until 3 October 2028. Download (119kB) | Request a copy
	Text 11221031_statement_of_authenticity.pdf Restricted to Repository staff only until 3 October 2028. Download (263kB) | Request a copy
	Text 11221031_publishing_agreement.pdf Restricted to Repository staff only until 3 October 2028. Download (161kB) | Request a copy
	Text 11221031_approval_sheet.pdf Restricted to Repository staff only until 3 October 2028. Download (249kB) | Request a copy
	Text 11221031_prefase.pdf Restricted to Repository staff only until 3 October 2028. Download (175kB) | Request a copy
	Text 11221031_abstract_id.pdf Restricted to Repository staff only until 3 October 2028. Download (204kB) | Request a copy
	Text 11221031_abstract_en.pdf Restricted to Repository staff only until 3 October 2028. Download (230kB) | Request a copy
	Text 11221031_table_of_content.pdf Restricted to Repository staff only until 3 October 2028. Download (220kB) | Request a copy
	Text 11221031_illustrations.pdf Restricted to Repository staff only until 3 October 2028. Download (181kB) | Request a copy
	Text 11221031_tables.pdf Restricted to Repository staff only until 3 October 2028. Download (182kB) | Request a copy
	Text 11221031_notations.pdf Restricted to Repository staff only until 3 October 2028. Download (210kB) | Request a copy
	Text 11221031_chapter_1.pdf Restricted to Repository staff only until 3 October 2028. Download (256kB) | Request a copy
	Text 11221031_chapter_2.pdf Restricted to Repository staff only until 3 October 2028. Download (465kB) | Request a copy
	Text 11221031_chapter_3.pdf Restricted to Repository staff only until 3 October 2028. Download (590kB) | Request a copy
	Text 11221031_chapter_4.pdf Restricted to Repository staff only until 3 October 2028. Download (652kB) | Request a copy
	Text 11221031_conclusions.pdf Restricted to Repository staff only until 3 October 2028. Download (158kB) | Request a copy
	Text 11221031_bibliography.pdf Restricted to Repository staff only until 3 October 2028. Download (287kB) | Request a copy
	Text 11221031_paper.pdf Restricted to Repository staff only until 3 October 2028. Download (581kB) | Request a copy
	Text 11221031_presentation.pdf Restricted to Repository staff only until 3 October 2028. Download (1MB) | Request a copy
	Text 11221031_Form. TA-020.pdf Restricted to Repository staff only until 3 October 2028. Download (53kB) | Request a copy

Abstract

Prediksi curah hujan harian yang akurat sangat penting untuk peringatan dini bencana hidrometeorologi, perencanaan infrastruktur, dan pengelolaan sumber daya air di daerah tropis. Studi ini mengusulkan Artificial Neural Network (ANN) yang dioptimalkan menggunakan Bayesian Optimization HyperBand (BOHB) untuk prediksi curah hujan harian di BMKG SAMS Sepinggan, Balikpapan, Indonesia. Dataset sebanyak 1.035 observasi harian (Januari 2023–Oktober 2025) dengan variabel meteorologi, termasuk suhu minimum (Tn), suhu maksimum (Tx), suhu rata-rata (Tavg), kelembaban relatif (RH_avg), dan durasi sinar matahari (ss), digunakan untuk memprediksi curah hujan harian (RR). Praproses data meliputi penghapusan kode BMKG yang tidak valid (8888/9999), penanganan outlier, interpolasi linier untuk nilai yang hilang, normalisasi MinMax, pembentukan jendela deret waktu dengan lookback optimal 5 hari, dan pemisahan data kronologis (70:15:15). Empat skenario eksperimental dievaluasi: ANN dasar tanpa optimasi, ANN dengan Bayesian Optimasi, ANN dengan Hyperband, dan ANN dengan BOHB. Hasil menunjukkan bahwa ANN yang dioptimalkan BOHB mencapai kinerja terbaik, dengan Test RMSE sebesar 6.937694 mm/hari dan MAE sebesar 5.868570 mm/hari, mengungguli semua pendekatan lainnya. Konfigurasi optimal terdiri dari learning rate sebesar 0.000169, batch size sebesar 64, jumlah hidden layer sebanyak 3 lapisan, jumlah neuron pada setiap lapisan sebesar 260, dropout rate sebesar 0.3735, optimizer yang digunakan yaitu Adagrad, serta fungsi aktivasi linier. Temuan ini menunjukkan bahwa optimasi hyperparameter berbasis BOHB secara signifikan meningkatkan kinerja ANN untuk prediksi deret waktu curah hujan tropis yang kompleks.

Item Type:	Thesis (Bachelor)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Jurusan Matematika dan Teknologi Informasi > Informatika
Depositing User:	Dilla Puspitasari
Date Deposited:	15 Jul 2026 05:24
Last Modified:	15 Jul 2026 05:24
URI:	http://repository.itk.ac.id/id/eprint/26909

Actions (login required)

View Item