ISWANTI, Arie and Isnanto, R. Rizal and Widodo, Catur Edi (2026) OPTIMASI PREDIKSI RISIKO PENYAKIT JANTUNG KORONER DENGAN ALGORITMA EXTREME LEARNING MACHINE: STUDI KASUS PASIEN RSUD DR. SOESELO. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text Cover aja.pdf Download (238kB)
	Text Cover dll.pdf Restricted to Repository staff only Download (803kB)
	Text BAB I.pdf Download (535kB)
	Text BAB II.pdf Download (1MB)
	Text BAB III.pdf Restricted to Repository staff only Download (765kB)
	Text BAB IV.pdf Restricted to Repository staff only Download (1MB)
	Text BAB V.pdf Restricted to Repository staff only Download (481kB)
	Text Daftar Pustaka.pdf Download (462kB)
	Text Lampiran.pdf Restricted to Repository staff only Download (311kB)

Abstract

Penyakit Jantung Koroner (PJK) tetap menjadi penyebab kematian utama secara global, sehingga sistem klasifikasi risiko yang akurat sangat krusial untuk intervensi klinis. Namun, dataset medis sering kali mengalami masalah ketidakseimbangan kelas (imbalanced data) yang memicu bias pada model prediktif. Penelitian ini bertujuan membangun model klasifikasi stadium risiko PJK (Kelas 0–4) menggunakan algoritma Extreme Learning Machine (ELM) yang dioptimalkan dengan teknik Synthetic Minority Over-sampling Technique (SMOTE). Penelitian menggunakan 521 data rekam medis pasien RSUD Dr. Soeselo Kabupaten Tegal tahun 2023–2024 yang dibagi menjadi 70% data latih dan 30% data uji. Optimasi dilakukan dengan menerapkan SMOTE pada data latih untuk menyeimbangkan komposisi kelas dan melakukan penalaan hyperparameter berupa penggunaan 100 hidden neurons dengan fungsi aktivasi sigmoid. Hasil pengujian pada 157 data uji menunjukkan model mencapai akurasi keseluruhan sebesar 82,17%. Analisis performa menunjukkan model sangat unggul dalam mengidentifikasi Kelas 0 (F1-Score 0,93) dan memberikan performa kuat pada Kelas 4 (F1-Score 0,84). Meskipun demikian, tantangan klasifikasi ditemukan pada Kelas 1 dan 3 dengan nilai recall sebesar 0,74 akibat karakteristik data yang tumpang tindih. Hasil ini membuktikan bahwa kombinasi SMOTE dan ELM efektif meningkatkan sensitivitas terhadap kelas minoritas dan dapat menjadi alat bantu keputusan klinis yang objektif.
Kata Kunci: Penyakit Jantung Koroner, Extreme Learning Machine (ELM), Prediksi Risiko, SMOTE, Klasifikasi Multi-kelas

Coronary Heart Disease (CHD) remains a leading cause of mortality globally, making accurate risk classification systems vital for clinical intervention. However, medical datasets frequently encounter class imbalance issues, which can lead to bias in predictive models. This study aims to develop a risk stage classification model (Classes 0–4) for CHD patients using the Extreme Learning Machine (ELM) algorithm, optimized with the Synthetic Minority Over-sampling Technique (SMOTE). The research utilized 521 medical records from patients at RSUD Dr. Soeselo Tegal Regency (2023–2024), partitioned into 70% training data and 30% testing data. Optimization was achieved by applying SMOTE to the training set to balance class distribution and performing hyperparameter tuning, specifically utilizing 100 hidden neurons with a sigmoid activation function. The testing phase on 157 unseen data points demonstrated that the model achieved an overall accuracy of 82.17%. Performance analysis revealed that the model excelled in identifying Class 0 (F1-Score 0.93) and showed strong performance for Class 4 (F1-Score 0.84). Nevertheless, classification challenges persisted in Classes 1 and 3, with a recall of 0.74 due to overlapping data characteristics. These results prove that the combination of SMOTE and ELM effectively enhances sensitivity toward minority classes and can serve as an objective clinical decision support tool.
Keywords: Coronary Heart Disease, Extreme Learning Machine (ELM), Risk Prediction, SMOTE, Multi-class Classification

Actions (login required)

View Item