JOURNAL OF SCIENCE AND SOCIAL RESEARCH
Vol. 9 No. 4 (2026): August 2026 (1)

PREDICTING MORTALITY FOR COVID-19 PATIENTS USING DATA MINING AND MACHINE LEARNING APPROACHES

Elvin Khoirunnisa (Bina Nusantara University)
Jin Xu (INESC TEC and University of Porto)



Article Info

Publish Date
31 Aug 2026

Abstract

Abstract: Since December 2019, the COVID-19 pandemic has spread worldwide, posing a grave danger to global public health and severely threatening healthcare systems. Therefore, early prediction of COVID-19 case severity is crucial for saving lives and optimizing healthcare logistics. In this study, we utilize a blood sample dataset from infected patients in Wuhan, China, and construct machine learning models to predict patient mortality risk. Given that the testing data includes only three biomarkers, the experiments are divided into two parts: an "all features" scenario and a "three features" scenario. Based on feature selection and model optimization, candidate models are configured with different feature subsets and evaluated using both single algorithms and ensemble models on the training data. The trained models are then evaluated on the testing data. The results demonstrate that ensemble models achieve superior performance compared to single models across both feature scenarios. Notably, ensemble models using the "three features" configuration achieve the highest overall performance among all experimental setups, despite relying on fewer features. Our top F1 score for the positive class (death) approaches the baseline results, confirming the effectiveness of our experimental framework. Keywords: Predictive model, COVID-19 mortality, data mining, feature selection   Abstrak: Sejak Desember 2019, wabah COVID-19 telah menyebar ke seluruh dunia, tidak hanya membahayakan nyawa dan kesehatan penduduk global serta mengancam sistem layanan kesehatan masyarakat secara serius. Oleh karena itu, prediksi dini mengenai tingkat keparahan kasus COVID-19 sangatlah penting untuk kehidupan masyarakat dan perencanaan logistik kesehatan. Dalam penelitian ini, kami memanfaatkan basis data sampel darah dari pasien yang terinfeksi di wilayah Wuhan, Tiongkok, dan membangun model pembelajaran mesin (machine learning) untuk memprediksi risiko kematian pasien tersebut. Mengingat data pengujian hanya mencakup tiga biomarker, eksperimen ini dibagi menjadi dua bagian: skenario "semua fitur" dan "tiga fitur". Berdasarkan pemilihan fitur dan model, berbagai model dikonfigurasi dengan rangkaian fitur yang berbeda, lalu menerapkan model tunggal maupun model ensemble pada data pelatihan. Selanjutnya, model yang telah dilatih tersebut diuji menggunakan data pengujian. Hasil penelitian menunjukkan bahwa model ensemble mencapai kinerja yang lebih baik dibandingkan model tunggal, baik pada skenario "semua fitur" maupun "tiga fitur". Hasil dari model ensemble pada skenario "tiga fitur" merupakan yang tertinggi di antara seluruh rangkaian eksperimen, meskipun menggunakan jumlah fitur yang lebih sedikit. Skor F1 tertinggi kami untuk sampel positif (death) mendekati hasil eksperimen dasar (baseline), yang menunjukkan efektivitas pendekatan eksperimental dalam penelitian ini. Kata kunci: Model prediktif, mortalitas COVID-19, penambangan data, pemilihan fitur

Copyrights © 2026






Journal Info

Abbrev

JSSR

Publisher

Subject

Computer Science & IT Economics, Econometrics & Finance Education Social Sciences

Description

Journal of Science and Social Research is accepts research works from academicians in their respective expertise of studies. Journal of Science and Social Research is platform to disclose the research abilities and promote quality and excellence of young researchers and experienced thoughts towards ...