To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
MML: Multimodal Machine Learning Strategies for the Prediction of Covid-19 Status and Spread |
|
Author |
Adebayo Nafisat Yemisi, Abiodun Esther Omolara, Olumide Owolabi, Oludare Isaac Abiodun |
| Citation |
Vol. 26 No. 4 pp. 172-196 |
|
Abstract |
Few years ago, the world was thrown into a sudden chaotic loop due to the emergence of COVID-19 virus. The virus rapidly spread across environments, villages, towns, cities, regions, countries and continents. Consequently, it culminated into a huge strain on healthcare systems, thereby creating a disproportionate sudden and poor response with respect to medical services. Years down the line, the pandemic continues to pose a significant threat to global health, social interactions, and the economy. Despite ongoing efforts, completely containing the spread of the virus has remained difficult due to the elusive nature of its origins, or probably given that it is a new outbreak of its kind. Indeed, it is perplexing to experts and researchers in the field on how to totally address the challenge. Predicting the status and spread of the COVID-19 virus is still difficult because of its ever-changing nature, including the variability in its occurrence and symptoms in humans, potentially related to its cause and origin. But now that machine learning techniques are performing well in prediction of unknown event. Therefore, this study employed multimodal machine learning (MML) strategies, utilizing dataset from the World Health Organization (WHO), National Centre for Disease Control (NCDC), and John Hopkins University. The data was preprocessed using one-hot encoding, mutual information, and robust statistical methods, with L2 regularization added to the neural network. The results showed that the neural network (NN) predicted global COVID-19 status with 96.2% accuracy and feature importance analysis with 96.4% accuracy. The decision tree model predicted COVID-19 status in Nigeria with 0.93 accuracy and an F1 score of 0.98, identifying Lagos State as the most at risk, with 104,285 incident cases and 1,142 cases treated. The study concludes that MML can reliably predict COVID-19 status and spread factors, with NN models showing high accuracy. The study recommends the use of neural networks for COVID-19 spread analysis but advises caution in long-term predictions. |
|
Keywords |
Covid-19, multimodal analysis of Covid-19, neural network, machine learning models. |
|
URL |
