Data Balancing Methods on Radiographic Image Classification on Unbalance Dataset
Abstract
Covid-19 is a disease caused by a corona virus infection that infects the victim's respiratory tract. Covid-19 disease has a high infectious ability and if treated too late can result in death. Covid-19 has been a problem faced by everyone in the world since the end of 2019. Fast and accurate detection can save many lives. This study aims to develop a predictive model of COVID-19 detection based on radiographic images using a machine learning model from 4 categories of health status, namely positive covid, normal, lung opacity sufferers, and viral pneumonia sufferers. Deep learning is based on ResNet50 and MobileNetV2 and trials of undersampling and oversampling data balancing methods, and uses a confusion matrix for the evaluation process of model results. The model with the highest performance achieves 95.58% accuracy in the multi-class classification. Also based on the findings, we provide results from using a different data balancing approach or not using one at all.
Keywords
classification, data balancing, deep neural network, machine learning, radiographic imagery
Full Text:
PDFDOI: https://doi.org/10.35793/jtek.11.1.2022.37186
Refbacks
- There are currently no refbacks.
Copyright (c) 2022 Joshua Axel Wuisan, Agustinus Jacobus, Sherwin Sompie
Published by Electrical Engineering Study Program, Sam Ratulangi University, Manado
Print-ISSN : 2301-8402 Electronic-ISSN: 2685-368X
Jurnal Teknik Elektro dan Komputer (JTEK) is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
INDEXED BY :