Analisis Prediktif Dinamika Penyebaran Kasus Demam Berdarah Menggunakan Dynamic Network Link Prediction Berbasis GAT-GRU

Ihsani, Shabrina Nur (2026) Analisis Prediktif Dinamika Penyebaran Kasus Demam Berdarah Menggunakan Dynamic Network Link Prediction Berbasis GAT-GRU. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026221002-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Demam Berdarah Dengue (DBD) merupakan penyakit menular yang penyebarannya dipengaruhi oleh interaksi spasial antarwilayah dan dinamika temporal sehingga memerlukan pendekatan prediksi yang mampu merepresentasikan keduanya. Pendekatan prediksi konvensional umumnya hanya berfokus pada jumlah kasus dan belum mampu memodelkan pola penyebaran antarwilayah secara eksplisit. Oleh karena itu, penelitian ini bertujuan membangun model prediksi dinamika penyebaran DBD di Kabupaten Malang menggunakan pendekatan Dynamic Network Link Prediction berbasis kombinasi GAT dan GRU. Data yang digunakan meliputi data insiden DBD harian, data cuaca, Angka Bebas Jentik (ABJ), kepadatan penduduk, dan batas administrasi kecamatan yang direpresentasikan sebagai graf dinamis. Eksperimen dilakukan dengan menguji variasi sequence length, penggunaan fitur lingkungan, bobot spasial berbasis Inverse Distance Weighting (IDW), rasio negative sampling, serta optimasi hyperparameter menggunakan grid search. Konfigurasi terbaik diperoleh pada sequence length 14 hari, penggunaan fitur lingkungan, rasio negative sampling 1:3, hidden dimension 64, dua attention head, dan satu lapisan GRU dengan nilai AUC 0,9699, accuracy 0,9675, precision 0,4246, recall 0,7180, dan F1-score 0,5336. Konfigurasi tersebut juga menghasilkan representasi jaringan penyebaran yang paling mendekati kondisi aktual pada periode normal, outbreak, maupun kasus rendah. Hasil penelitian menunjukkan bahwa model GAT-GRU mampu mempelajari hubungan spasial dan dinamika temporal secara efektif sehingga berpotensi mendukung sistem peringatan dini dan pengambilan keputusan dalam pengendalian penyebaran DBD.
================================================================================================================================
Dengue Hemorrhagic Fever (DHF) is an infectious disease whose transmission is influenced by spatial interactions among regions and temporal dynamics, requiring a predictive approach capable of representing both aspects simultaneously. Conventional prediction approaches generally focus only on the number of cases and are unable to explicitly model interregional transmission patterns. Therefore, this study aims to develop a predictive model for the dynamics of DHF transmission in Malang Regency using a Dynamic Network Link Prediction approach based on the combination of Graph Attention Network (GAT) and Gated Recurrent Unit (GRU). The data used include daily DHF incidence data, weather data, Larvae-Free Index (ABJ), population density, and subdistrict administrative boundary data, which are represented as a dynamic graph. The experiments evaluate variations in sequence length, the use of weather and social features, Inverse Distance Weighting (IDW)-based spatial edge weights, negative sampling ratios, and hyperparameter optimization using grid search. The best configuration was obtained using a 14-day sequence length, weather and social features, a 1:3 negative sampling ratio, a hidden dimension of 64, two attention heads, and one GRU layer, achieving an AUC of 0.9699, accuracy of 0.9675, precision of 0.4246, recall of 0.7180, and an F1-score of 0.5336. This configuration also produced a transmission network representation that most closely resembled the actual conditions during normal, outbreak, and low-incidence periods. The results demonstrate that the proposed GAT-GRU model effectively captures both spatial relationships and temporal dynamics, making it a promising approach for supporting early warning systems and decision-making in DHF transmission control.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Demam Berdarah Dengue, Prediksi Spasio-Temporal, Dynamic Network Link Prediction, Graph Attention Network, Gated Recurrent Unit. Dengue Hemorrhagic Fever, Spatio-Temporal Prediction, Dynamic Network Link Prediction, Graph Attention Network, Gated Recurrent Unit.
Subjects:	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:	Shabrina Nur Ihsani
Date Deposited:	30 Jul 2026 06:09
Last Modified:	30 Jul 2026 06:09
URI:	http://repository.its.ac.id/id/eprint/139761

Actions (login required)

View Item