Vol. 12, Issue 2 (2023)

Author(s):
Ambreen Hamadani, Onyijen Ojei Harrison, Nazir A Ganai, Israel Ehizuelen Ebhohimen, Oluwatobi Samuel Awe, Celestine Uche Agwi and J Bashir

Abstract:
The COVID-19 pandemic had a great impact on the world. It wreaked a great havoc around the globe impacting every human on the face of this planet in one way or the other. The unprecedented challenges that the world faced demonstrated the need for big data and its mining for making futuristic predictions so that disasters such as this could be averted, especially when human lives are at stake. The present study was therefore undertaken to harness the potential of artificial intelligence algorithms for drawing useful inferences from the COVID 19 dataset by “Our World in Data”. Appropriate data preparation was done using data imputation, encoding, and normalization. Feature selection was performed on the 67 initial features. Ordinary least squares and artificial neural networks (ANN) were used in this research. Convolutional Neural Networks (CNN) were also compared with the feed-forward back propagation algorithm for the current dataset. Our results indicate that the total cases, total tests, total vaccinations, and diabetes prevalence had the highest feature scores of 10983.38, 6636.90, 5118.28, and 2150.80 respectively. The coefficient of determination value for the regression equation was 0.972 indicating a good fit. For the ANN model with only one feature i.e., total vaccination, the correlation coefficient was 0.865 going on to show that vaccination was an important factor in preventing deaths. Our results also indicate that both the ANN and CNN had high prediction correlations (0.99), but the CNN had a lower error rate for the same. Additionally, the high feature scores of factors like total cases and total vaccinations influence the death rates greatly. Therefore, preventive measures like social distancing, masking up and vaccination could potentially reduce the death rate associated with COVID-19. We conclude that the machine learning models developed were accurate based on the correlation coefficient and could be used for drawing useful patterns regarding deaths during this great pandemic of the Century.