Gagal ginjal kronis merupakan salah satu penyakit serius yang dapat menyebabkan kematian jika tidak terdeteksi dan ditangani secara dini. Penelitian ini bertujuan memprediksi risiko kematian pada pasien gagal ginjal kronis menggunakan metode ensemble learning, yaitu random forest dan voting classifier (hard voting dan soft voting). Voting classifier digunakan untuk menggabungkan prediksi dari beberapa model klasifikasi tunggal, di mana hard voting mengambil keputusan berdasarkan suara terbanyak, sedangkan soft voting mempertimbangkan rata-rata probabilitas prediksi. Data yang digunakan pada penelitian ini merupakan data sekunder dari RSUD Dr. Saiful Anwar, Kota Malang. Proporsi pasien rawat inap yang pulang dalam kondisi meninggal lebih kecil dibanding kondisi tidak meninggal. Kondisi data tidak seimbang ini menyebabkan model cenderung bias terhadap kelas mayoritas. Untuk mengatasi hal tersebut, synthetic minority over-sampling technique (SMOTE) diterapkan guna menyeimbangkan distribusi kelas. Selain itu, random forest dipilih karena kemampuannya dalam menangani ketidakimbangan data melalui pembobotan pada pohon-pohon keputusan, sehingga mengurangi bias terhadap kelas mayoritas. Evaluasi performa model dilakukan menggunakan metrik akurasi, presisi, dan recall. Hasil evaluasi menunjukkan bahwa random forest memberikan kinerja terbaik dengan akurasi 77%, presisi 36%, dan recall 60%, mengungguli hard voting dan soft voting. Penggunaan random forest dan SMOTE terbukti meningkatkan prediksi pada kelas minoritas, yang sangat penting dalam mendeteksi pasien berisiko kematian tinggi. Pendekatan ini dapat membantu dalam deteksi dini dan pengelolaan yang lebih baik terhadap pasien gagal ginjal kronis, sehingga berpotensi menurunkan angka kematian akibat penyakit ini. Abstract Chronic kidney disease (CKD) is a life-threatening condition that can lead to fatal outcomes if not diagnosed and treated promptly. This study aims to forecast mortality risk in CKD patients using ensemble learning techniques, including random forest an d voting classifier (hard voting and soft voting). The voting classifier combines predictions from various single classification models, with hard voting selecting outcomes based on majority decisions, while soft voting averages prediction probabilities. The data used in this study is secondary data from RSUD Dr. Saiful Anwar, Malang City. The proportion of hospitalized patients who were discharged in a deceased condition is smaller than those who were discharged alive. This imbalance in the data causes the model to be biased toward the majority class. However, models tend to favor the majority class when dealing with imbalanced data. To mitigate this, the synthetic minority over-sampling technique (SMOTE) was applied to balance the class distribution. Random forest was also selected for its ability to manage data imbalance through weighted decision trees, reducing bias toward the majority class. Model performance was evaluated using metrics such as accuracy, precision, and recall. Results indicated that random forest outperformed hard voting and soft voting, achieving 77% accuracy, 36% precision, and 60% recall. The combination of random forest and SMOTE significantly enhanced the prediction of minority class outcomes, which is essential for identifying high-risk patients. This method has the potential to support early detection and improved management of CKD patients, thus reducing mortality rates associated with the disease.
English
Gagal ginjal kronis merupakan salah satu penyakit serius yang dapat menyebabkan kematian jika tidak terdeteksi dan ditangani secara dini. Penelitian ini bertujuan memprediksi risiko kematian pada pasien gagal ginjal kronis menggunakan metode ensemble learning, yaitu random forest dan voting classifier (hard voting dan soft voting). Voting classifier digunakan untuk menggabungkan prediksi dari beberapa model klasifikasi tunggal, di mana hard voting mengambil keputusan berdasarkan suara terbanyak, sedangkan soft voting mempertimbangkan rata-rata probabilitas prediksi. Data yang digunakan pada penelitian ini merupakan data sekunder dari RSUD Dr. Saiful Anwar, Kota Malang. Proporsi pasien rawat inap yang pulang dalam kondisi meninggal lebih kecil dibanding kondisi tidak meninggal. Kondisi data tidak seimbang ini menyebabkan model cenderung bias terhadap kelas mayoritas. Untuk mengatasi hal tersebut, synthetic minority over-sampling technique (SMOTE) diterapkan guna menyeimbangkan distribusi kelas. Selain itu, random forest dipilih karena kemampuannya dalam menangani ketidakimbangan data melalui pembobotan pada pohon-pohon keputusan, sehingga mengurangi bias terhadap kelas mayoritas. Evaluasi performa model dilakukan menggunakan metrik akurasi, presisi, dan recall. Hasil evaluasi menunjukkan bahwa random forest memberikan kinerja terbaik dengan akurasi 77%, presisi 36%, dan recall 60%, mengungguli hard voting dan soft voting. Penggunaan random forest dan SMOTE terbukti meningkatkan prediksi pada kelas minoritas, yang sangat penting dalam mendeteksi pasien berisiko kematian tinggi. Pendekatan ini dapat membantu dalam deteksi dini dan pengelolaan yang lebih baik terhadap pasien gagal ginjal kronis, sehingga berpotensi menurunkan angka kematian akibat penyakit ini. Abstract Chronic kidney disease (CKD) is a life-threatening condition that can lead to fatal outcomes if not diagnosed and treated promptly. This study aims to forecast mortality risk in CKD patients using ensemble learning techniques, including random forest an d voting classifier (hard voting and soft voting). The voting classifier combines predictions from various single classification models, with hard voting selecting outcomes based on majority decisions, while soft voting averages prediction probabilities. The data used in this study is secondary data from RSUD Dr. Saiful Anwar, Malang City. The proportion of hospitalized patients who were discharged in a deceased condition is smaller than those who were discharged alive. This imbalance in the data causes the model to be biased toward the majority class. However, models tend to favor the majority class when dealing with imbalanced data. To mitigate this, the synthetic minority over-sampling technique (SMOTE) was applied to balance the class distribution. Random forest was also selected for its ability to manage data imbalance through weighted decision trees, reducing bias toward the majority class. Model performance was evaluated using metrics such as accuracy, precision, and recall. Results indicated that random forest outperformed hard voting and soft voting, achieving 77% accuracy, 36% precision, and 60% recall. The combination of random forest and SMOTE significantly enhanced the prediction of minority class outcomes, which is essential for identifying high-risk patients. This method has the potential to support early detection and improved management of CKD patients, thus reducing mortality rates associated with the disease.
Affiliations come from OpenAlex; journal OAI-PMH metadata carries none. Authors without an OpenAlex ID are matched by name, which can merge or split people.