OPTIMIZATION OF LONG SHORT-TERM MEMORY ALGORITHM USING PARTICLE SWARM OPTIMIZATION FOR RAINFALL PREDICTION IN PASURUAN REGENCY

Fisena, Muhammad Reyhan Dwi (2026) OPTIMIZATION OF LONG SHORT-TERM MEMORY ALGORITHM USING PARTICLE SWARM OPTIMIZATION FOR RAINFALL PREDICTION IN PASURUAN REGENCY. Undergraduate thesis, Universitas Pembangunan Nasional "Veteran" Jawa TImur.

	Text (Cover) 22081010198.-cover.pdf Download (1MB)
	Text (BAB I) 22081010198.-bab1.pdf Download (288kB)
	Text (BAB II) 22081010198.bab2.pdf Restricted to Repository staff only until 25 June 2028. Download (488kB)
	Text (BAB III) 22081010198.-bab3.pdf Restricted to Repository staff only until 25 June 2028. Download (906kB)
	Text (BAB IV) 22081010198.-bab4.pdf Restricted to Repository staff only until 25 June 2028. Download (3MB)
	Text (BAB V) 22081010198.-bab5.pdf Download (188kB)
	Text (Daftar Pustaka) 22081010198.-daftarpustaka.pdf Download (134kB)
	Text (Lampiran) 22081010198.-lampiran.pdf Restricted to Repository staff only Download (241kB)

Abstract

Pasuruan Regency is a region in East Java that is highly vulnerable to hydrometeorological disasters, a condition exacerbated by increasing rainfall intensity and the limitations of existing forecasting systems. This study develops a daily rainfall prediction model using the Long Short-Term Memory (LSTM) algorithm optimized with Particle Swarm Optimization (PSO). The dataset was obtained from the Pasuruan Geophysical Station of the BMKG, consisting of daily temperature, humidity, and rainfall data from January 1, 2021, to December 31, 2025 (1,815 records). Data preprocessing included missing value handling using Random Forest Imputation, normalization using the Min-Max Scaler, and transformation through the sliding window technique. PSO was employed to optimize the LSTM hyperparameters through ten experimental scenarios with variations in inertia weight, cognitive coefficient, and social coefficient. The best-performing model was obtained in the LSTM-PSO7 scenario, with a configuration of 192 neurons in the first layer, 103 neurons in the second layer, a learning rate of 0.009126, and a batch size of 16. Evaluation results showed that LSTM-PSO7 achieved an RMSE of 18.023, MAE of 12.413, and NSE of 0.134, outperforming the standard LSTM, GRU, ARIMA, and TES models. The final model was implemented as an interactive Streamlit-based web application to support disaster mitigation decision-making in Pasuruan Regency.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Puspaningrum, Eva Yulia	19890705 202121 2 002	evapuspaningrum.if@upnjatim.ac.id
Thesis advisor	Rakhmadi, Ardhon	19910805 202406 1 002	ardhon.rakhmadi.fasilkom@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QA Mathematics > QA76.6 Computer Programming

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Muhammad Reyhan Dwi Fisena

Date Deposited:

25 Jun 2026 04:48

Last Modified:

25 Jun 2026 06:45

URI:

https://repository.upnjatim.ac.id/id/eprint/54202