Analisis Prediktif Penyebaran Demam Berdarah di Daerah Endemis dengan Pendekatan EvolveGCN: Studi Kasus Kabupaten Malang, Indonesia

Meilani, Maulidiya (2025) Analisis Prediktif Penyebaran Demam Berdarah di Daerah Endemis dengan Pendekatan EvolveGCN: Studi Kasus Kabupaten Malang, Indonesia. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211121-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (3MB) | Request a copy

Abstract

Demam berdarah (DBD) merupakan masalah kesehatan global yang signifikan, terutama di daerah tropis dan subtropis seperti Indonesia. Kabupaten Malang, sebagai salah satu daerah endemis, sering mengalami peningkatan kasus DBD. Penelitian ini bertujuan untuk mengembangkan model prediktif yang efektif dalam memperkirakan penyebaran DBD dengan menggunakan pendekatan EvolveGCN. Gabungan antara Graph Convolutional Network (GCN) dengan Recurrent Neural Network (RNN) ini mampu menangkap dinamika spasial-temporal data. Metodologi penelitian meliputi studi literatur dan identifikasi permasalahan, pengumpulan dan pra-pemrosesan data historis kasus DBD serta faktor-faktor lingkungan, geografis, dan sosial, setelah itu graf dinamis dibentuk, dan pelatihan model EvolveGCN. Hasil penelitian menunjukkan bahwa kedua varian model EvolveGCN-O dan EvolveGCN-H mencapai nilai AUC 91% pada data pengujian, dengan recall 91-92%, meskipun precision hanya mencapai 18%. Konfigurasi optimal diperoleh menggunakan 7 output channels dan window size 21 hari. Model menunjukkan kecenderungan over-prediction, mengindikasikan perlunya penyempurnaan dalam presisi prediksi. Untuk pengembangan ke depan, disarankan untuk menerapkan teknik regularisasi, mengeksplorasi window size yang lebih besar, menambahkan fitur-fitur baru yang lebih informatif, dan mengembangkan pendekatan khusus untuk menangani karakteristik sparse graph. Model ini diharapkan dapat membantu pihak berwenang dalam merancang strategi pencegahan dan pengendalian DBD yang lebih efektif.
====================================================================================================================================
Dengue fever (DF) represents a significant global health concern, particularly in tropical and subtropical regions like Indonesia. Malang Regency, as one of the endemic areas, frequently experiences increased cases of dengue fever. This research aims to develop an effective predictive model for estimating dengue fever spread using the EvolveGCN approach. The combination of Graph Convolutional Network (GCN) and Recurrent Neural Network (RNN) can capture the spatial-temporal dynamics of data. The research methodology includes literature study and problem identification, collection and pre-processing of historical dengue fever case data along with environmental, geographical, and social factors, followed by dynamic graph formation and EvolveGCN model training. Results show that both EvolveGCN-O and EvolveGCN-H model variants achieved an AUC value of 91% on test data, with 91-92% recall, although precision only reached 18%. Optimal configuration was achieved using 7 output channels and a 21-day window size. The model showed a tendency for over-prediction, indicating the need for improvement in prediction precision. For future development, it is recommended to implement regularization techniques, explore larger window sizes, add more informative features, and develop specific approaches to handle sparse graph characteristics. This model is expected to assist authorities in designing more effective dengue fever prevention and control strategies.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kata kunci: Demam Berdarah, Jaringan Dinamis Berarah, Analisis Spatio-Temporal, Link Prediction, EvolveGCN Keywords: Dengue Fever, Directed Dynamic Networks, Spatio-Temporal Analysis, Link Prediction, EvolveGCN
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Maulidiya Meilani
Date Deposited:	01 Feb 2025 15:09
Last Modified:	01 Feb 2025 15:09
URI:	http://repository.its.ac.id/id/eprint/117630

Actions (login required)

View Item