Penerapan Model BiLSTMAttention dengan Optimasi Moth Flame untuk Prediksi Kadar Particulate Matter (PM10)

Kurniawan, Mohammad Lucky (2025) Penerapan Model BiLSTMAttention dengan Optimasi Moth Flame untuk Prediksi Kadar Particulate Matter (PM10). Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 21081010206.-cover.pdf Download (963kB)
	Text (Bab 1) 21081010206.-bab1.pdf Download (224kB)
	Text (Bab 2) 21081010206.-bab2.pdf Restricted to Repository staff only until 8 December 2028. Download (853kB)
	Text (Bab 3) 21081010206.-bab3.pdf Restricted to Repository staff only until 8 December 2028. Download (665kB)
	Text (Bab 4) 21081010206.-bab4.pdf Restricted to Repository staff only until 8 December 2028. Download (3MB)
	Text (Bab 5) 21081010206.-bab5.pdf Download (183kB)
	Text (Daftar pustaka) 21081010206.-daftarpustaka.pdf Download (182kB)
	Text (Lampiran) 21081010206.-lampiran.pdf Restricted to Repository staff only until 2028. Download (314kB)

Abstract

This research aims to predict the concentration of Particulate Matter (PM₁₀) as one of the main indicators of air pollution by implementing the Bidirectional Long Short-Term Memory (BiLSTM) model, which is enhanced with an Attention Mechanism and optimized using the Moth Flame Optimization (MFO) algorithm. The research methodology consists of eight main stages, namely data collection of air quality from Satu Data Jakarta, integration with meteorological data from BMKG, data preprocessing (including cleaning, smoothing, normalization, and sequence formation), model development of BiLSTM-Attention, hyperparameter optimization using MFO, model training, testing scenarios, and model evaluation. The BiLSTM model is utilized to capture bidirectional temporal dependencies within the time-series data, while the Attention Mechanism focuses on emphasizing the most relevant features that influence prediction outcomes. The MFO algorithm is employed to optimize crucial hyperparameters such as the number of neurons, learning rate, and dropout rate in order to obtain the best-performing configuration. The experimental results indicate that the BiLSTM-Attention-MFO model achieves the most optimal performance, with MAE of 2.2788, RMSE of 2.9774, MAPE of 5.01%, and R² of 0.9316. These findings demonstrate that the integration of the MFO algorithm effectively enhances the accuracy and stability of the BiLSTM-Attention model. Therefore, this model can be effectively utilized for predicting PM₁₀ concentrations and can serve as a decision-support tool for early warning systems and air pollution mitigation policies.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Sari, Anggraini Puspita	NIDN0716088605	anggraini.puspita.if@upnjatim.ac.id
Thesis advisor	Puspaningrum, Eva Yulia	NIDN0005078908	evapuspaningrum.if@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.6 Computer Programming
T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Kurniawan Mohammad Lucky

Date Deposited:

08 Dec 2025 03:05

Last Modified:

08 Dec 2025 03:05

URI:

https://repository.upnjatim.ac.id/id/eprint/48202