Imputasi kNN pada Hybrid Support Vector Machine untuk Kasus Klasifikasi Tingkat Keparahan COVID-19

Jalil, Sayyid Nur Cahyo Abdul (2023) Imputasi kNN pada Hybrid Support Vector Machine untuk Kasus Klasifikasi Tingkat Keparahan COVID-19. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 06211940000089-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (3MB) | Request a copy

Abstract

COVID-19 merupakan penyakit infeksi pada organ-organ pernapasan yang disebabkan oleh virus SARS-CoV-2 dan telah menyebar ke seluruh dunia sejak 2019. Dalam mengurangi akibat yang mungkin disebabkan oleh COVID-19, diperlukan diagnosis yang lebih akurat dalam membedakan pasien COVID-19 berdasarkan tingkat keparahannya sehingga dapat dijadikan sebagai early warning dalam memberikan perawatan yang tepat pada pasien. Pencatatan rekam medis pasien dapat dijadikan prediktor dalam menduga tingkat keparahan pasien. Pada permasalahan klasifikasi di bidang medis, terdapat kemungkinan adanya missing value dikarenakan rekam medis yang tidak tercatat sepenuhnya serta terdapat berbagai noise, outlier dan minimnya data tersedia akibat kesalahan pencatatan data rekam medis yang dapat menyebabkan overfitting. Penelitian ini bertujuan mengembangkan model prediksi yang akurat dan dapat menangani overfitting menggunakan Hybrid Random Forest-Support Vector Machine (RF-SVM) serta mengatasi missing value yang terdapat pada data rekam medis pasien COVID-19 menggunakan imputasi kNN. Diharapkan penelitian ini mampu meningkatkan performa prediktif dan mengatasi overfitting dengan model prediktif hybrid RF-SVM serta mengatasi permasalahan missing value pada data rekam medis dengan metode imputasi pada studi kasus pasien COVID-19 di RS UNAIR sehingga dapat membantu pihak medis untuk memberikan treatment yang lebih sesuai pada pasien COVID-19. Hasil analisis pada penelitian ini menunjukkan bahwa sebagian besar pasien yang terjangkit COVID-19 menunjukkan risiko yang rendah serta memiliki karakteristik pada tingkat keparahan rendah/sedang antara lain pasien berusia muda, tidak pernah menderita diabetes dan hipertensi, laju pernafasan normal/rendah, saturasi oksigen tinggi, dan tidak menunjukkan gejala sesak nafas. Penggunaan imputasi kNN menunjukkan bahwa metode ini mampu menghasilkan performa prediksi yang lebih besar dibandingkan metode penanganan missing value lainnya. Penggunaan metode hybrid RF-SVM menunjukkan bahwa metode ini mampu secara efektif mengurangi overfitting dan meningkatkan kemampuan generalisasi model. Hasil permodelan menggunakan imputasi kNN dan hybrid Random Forest – Support Vector Machine menunjukkan performa akurasi mencapai 97,6% dalam memprediksi data baru.
=================================================================================================================================
COVID-19 is an infectious disease of the respiratory organs caused by the SARS-CoV-2 virus and has spread throughout the world since 2019. In reducing the consequences that may be caused by COVID-19, a more accurate diagnosis is needed in differentiating COVID-19 patients. 19 based on the level of severity so that it can be used as an early warning in providing appropriate care to patients. The recording of the patient's medical record can be used as a predictor in predicting the severity of the patient. In classification problems in the medical field, there is a possibility of missing values due to medical records that are not fully recorded and there are various noises, outliers, and lack of available data due to errors in recording medical record data which can cause overfitting. This study aims to develop a predictive model that is accurate and can handle overfitting using Hybrid Random Forest-Support Vector Machine (RF-SVM) and overcoming missing values contained in the medical record data of COVID-19 patients using kNN imputation. It is hoped that this research will be able to improve predictive performance and overcome overfitting with the RF-SVM hybrid predictive model as well as overcome the problem of missing values in medical record data using the imputation method in case studies of COVID-19 patients at UNAIR Hospital so that it can help medical parties to provide more appropriate treatment. in COVID-19 patients. The results of the analysis in this study showed that the majority of patients who contracted COVID-19 showed mild/moderate risk and had characteristics at mild/moderate severity, including patients who were young, never had diabetes and hypertension, normal/low respiratory rate, high oxygen saturation, and did not show symptoms of shortness of breath. The use of kNN imputation shows that this method is capable of producing greater predictive performance than other missing value handling methods. The use of the RF-SVM hybrid method shows that this method is able to effectively reduce overfitting and increase the generalization ability of the model. Modeling results using kNN imputation and hybrid Random Forest – Support Vector Machine show an accuracy performance of up to 97.6% in predicting new data.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	COVID-19, Hybrid Support Vector Machine, Imputasi kNN, Klasifikasi, Classification, kNN Imputation
Subjects:	Q Science > Q Science (General) Q Science > QA Mathematics
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Sayyid Nur Cahyo Abdul Jalil
Date Deposited:	15 Sep 2023 07:06
Last Modified:	15 Sep 2023 07:06
URI:	http://repository.its.ac.id/id/eprint/104554

Actions (login required)

View Item