Prediksi Hubungan Jaringan Dinamis Berarah pada Persebaran Demam Berdarah Dengue Menggunakan GAT-LSTM

Hanny, Hanny (2025) Prediksi Hubungan Jaringan Dinamis Berarah pada Persebaran Demam Berdarah Dengue Menggunakan GAT-LSTM. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211070-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Penyebaran penyakit Demam Berdarah Dengue (DBD) di Indonesia merupakan masalah kesehatan yang signifikan, terutama karena kurangnya sistem prediksi yang mampu mengidentifikasikan pola penyebaran secara efektif. Tugas akhir ini mengusulkan model prediksi hubungan dalam jaringan dinamis berarah menggunakan Graph Attention Network-Long Short-Term Memory (GAT-LSTM). Model ini memanfaatkan keunggulan GAT dalam menangkap informasi spasial antar node dan LSTM dalam memahami pola temporal pada jaringan. Data yang digunakan mencakup kejadian DBD, angka bebas jentik, karakteristik lokasi, dan data cuaca yang diperoleh dari berbagai instansi resmi untuk periode 2019–2023. Model GAT-LSTM akan diintegrasikan dengan decoder untuk menghasilkan matriks probabilitas yang merepresentasikan hubungan antar node. Evaluasi dilakukan menggunakan metrik Area Under the Curve (AUC) dan Error Rate untuk menilai performa model. Hasil eksperimen menunjukkan bahwa model mencapai skor AUC yang tinggi yaitu 93,15%, namun model cenderung melakukan prediksi berlebih terhadap hubungan yang tidak terjadi akibat ketidakseimbangan kelas dari data kasus insiden DBD yang ada. Terdapat indikasi overfitting dari perbedaan performa AUC antara data latih (95,20%) dan validasi (82,79%). Kombinasi terbaik diperoleh pada konfigurasi 1 attention head, 64 hidden channel, 2 layer LSTM, dan 15 window size. Temuan ini menunjukkan potensi pendekatan GAT-LSTM dalam mendukung pengambilan keputusan untuk pengendalian dan pencegahan DBD di Kabupaten Malang.
===========================================================================================================================================
The spread of Dengue Fever (DF) in Indonesia is a significant health issue, primarily due to the lack of predictive systems capable of effectively identifying the patterns of disease transmission. This final project proposes a link prediction model in dynamic directed networks using Graph Attention Network-Long Short-Term Memory (GAT-LSTM). The model leverages GAT's ability to capture spatial information between nodes and LSTM's capability to understand temporal patterns in the network. The data used includes DHF cases, larva-free index, location characteristics, and weather data obtained from various official agencies for the 2019–2023 period. The GAT-LSTM model will be integrated with a decoder to produce a probability matrix representing the connections between nodes. Evaluation is conducted using Area Under the Curve (AUC) and Error Rate metrics to assess model performance. Experimental results indicate that the model achieves a high AUC score of 93.15%, although it tends to overpredict connections that do not actually occur, likely due to class imbalance from the dataset of dengue fever incident. Signs of overfitting are observed from the performance gap of AUC between training (95,20%) and validation (82,79%) data. The best configuration was obtained using 1 attention head, 64 hidden channels, 2 LSTM layers, and a window size of 15. These findings highlight the potential of the GAT-LSTM approach to support decision-making in the control and prevention of dengue outbreaks in Malang Regency.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Demam Berdarah Dengue, Link Prediction, Graph Attention Network, Long Short-Term Memory, Jaringan Dinamis Berarah, Dengue Fever, Dynamic Directed Network
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA166 Graph theory Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Hanny .
Date Deposited:	18 Jul 2025 03:50
Last Modified:	18 Jul 2025 03:50
URI:	http://repository.its.ac.id/id/eprint/119998

Actions (login required)

View Item