Abstract
Timely diagnosis of COVID-19 is crucial to mitigate the risk of virus transmission. Traditional diagnostic methods, such as medical laboratory and antigen tests, while effective, are not always easily accessible. This study proposes an innovative approach to detect COVID-19 promptly using Artificial Neural Networks (ANN), eliminating the need for laboratory tests. By analyzing an individual's current symptoms, the ANN serves as a powerful tool for early diagnosis. The dataset employed in this research was sourced from Kaggle, specifically the COVID-19 presence and symptoms dataset. To enhance data pre-processing and hyperparameter tuning, GridSearchCV was utilized, incorporating 10-fold cross-validation. The optimal configuration, derived from these procedures, facilitated the construction of an effective prediction model using ANN. The findings reveal that hidden layer sizes of (100,), (50, 100, 50), and (50, 50, 50), coupled with relu and tanh activation functions, adam solver, alpha values of 0.05 and 0.0001, and adaptive or constant learning rates, collectively achieved the highest algorithm performance. Employing this optimal configuration, the ANN-based prediction model demonstrated an impressive 98.84% accuracy, 98.69% specificity, 100% sensitivity, and a 98.84% ROC curve. This developed prediction model holds the potential to revolutionize COVID-19 detection by enabling real-time identification of the disease without the reliance on laboratory tests. Applications utilizing this model could significantly contribute to early intervention and prevention strategies, ultimately reducing the spread of the virus in the community.
Author Information
Charlyn V. Rosales, Bulacan State University, Philippines
Paper Information
Conference: SEACE2024
Stream: Innovation & Technology
This paper is part of the SEACE2024 Conference Proceedings (View)
Full Paper
View / Download the full paper in a new tab/window
To cite this article:
Rosales C. (2024) Early Diagnosis Prediction From COVID-19 Symptoms Using ANN-Based Machine Learning Method ISSN: 2435-5240 The Southeast Asian Conference on Education 2024: Official Conference Proceedings https://doi.org/10.22492/issn.2435-5240.2024.29
To link to this article: https://doi.org/10.22492/issn.2435-5240.2024.29
Comments
Powered by WP LinkPress