FANIA, Dea and Waspada, Indra and Wibawa, Helmie Arif (2026) PENGEMBANGAN MODEL PREDIKSI PENYAKIT KARDIOVASKULAR BERBASIS RANDOM FOREST DENGAN ANALISIS SHAP PADA INTEGRASI DATA KLINIS MULTISUMBER. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text cover aja.pdf Download (193kB)
	Text a. Cover Tesis Dea Fania.pdf Restricted to Repository staff only Download (2MB)
	Text b. Bab I Tesis Dea Fania.pdf Download (326kB)
	Text c. Bab II Tesis Dea Fania.pdf Download (1MB)
	Text d. Bab III Tesis Dea Fania.pdf Restricted to Repository staff only Download (670kB)
	Text e. Bab IV Tesis Dea Fania.pdf Restricted to Repository staff only Download (1MB)
	Text F. Bab V Tesis Dea Fania.pdf Restricted to Repository staff only Download (301kB)
	Text G. Daftar Pustaka Tesis Dea Fania.pdf Download (306kB)

Abstract

Penyakit kardiovaskular (PKV) merupakan salah satu penyebab utama kematian di Indonesia sehingga diperlukan upaya pencegahan yang efektif, salah satunya melalui pengembangan sistem prediksi risiko berbasis data kesehatan masyarakat Indonesia. Tantangan utama dalam pengembangan sistem prediksi PKV adalah keterbatasan jumlah data medis lokal yang mampu merepresentasikan kondisi masyarakat di Indonesia. Penelitian ini bertujuan membangun model prediksi risiko PKV menggunakan algoritma Random Forest melalui integrasi dua sumber data, yaitu data rekam medis pasien poli jantung RSUD M. Yunus Kota Bengkulu sebagai data private dan dataset PKV dari Kaggle sebagai data public. Penelitian dilakukan melalui tiga eksperimen, yaitu Eksperimen Pengurangan Atribut, Eksperimen Perbandingan Performa Model Data Public dan Data Private serta Eksperimen Penggabungan Data. Analisis kontribusi setiap atribut dilakukan menggunakan metode Shapley Additive Explanations (SHAP). Hasil menunjukkan bahwa Eksperimen Penggabungan Data Skenario 3 menghasilkan performa model terbaik dari seluruh eksperimen dengan akurasi 73,57%, recall 81,44%, dan F1-score 77,06%. Analisis SHAP mengidentifikasi tekanan darah dan usia merupakan faktor paling berpengaruh dalam prediksi risiko PKV. Hasil dari seluruh Eksperimen menegaskan bahwa integrasi data private yang terbatas dengan data public mampu meningkatkan performa model. Selain itu, penggunaan SHAP memberikan panduan praktis untuk pengembangan model prediksi yang interpretatif secara klinis pada keterbatasan data private.
Kata Kunci : Penyakit Kardiovaskular, Random Forest, SHAP, Integrasi Data, Sistem Prediksi Risiko

Cardiovascular disease (CVD) is one of the leading causes of mortality in Indonesia, highlighting the need for effective prevention strategies, including the development of risk prediction systems based on population health data. A major challenge in developing such systems is the limited availability of local medical data that adequately represent the Indonesian population. This study aims to develop a CVD risk prediction model using the Random Forest algorithm by integrating two data sources: private medical record data from cardiology outpatients at RSUD M. Yunus Bengkulu and a public CVD dataset from Kaggle. The research was conducted through three experiments: the Attribute Reduction Experiment, the Model Performance Comparison between Public and Private Data Experiment, and the Data Integration Experiment. The contribution of each attribute was analyzed using Shapley Additive Explanations (SHAP). The results show that the best model achieved the highest performance among all experiments, with an accuracy of 73,57%, recall of 81,44%, and an F1-Score of 77,06%. SHAP analysis identified blood pressure and age as the most influential factors in predicting CVD risk. Overall, the findings confirm that integrating limited private data with public datasets can improve model performance. In addition, the use of SHAP provides practical guidance for developing clinically interpretable prediction models in settings with limited private data.
Keywords : Cardiovascular Disease, Random Forest, SHAP, Data Integration, Risk Prediction System

Actions (login required)

View Item