The ongoing COVID-19 pandemic has driven the attention of researchers for advanced and adaptable methods in disease detection. This research paper proposes a novel multi-agent-based model for detecting COVID-19 among patients with high accuracy.
Methods:Initially, all the necessary information is attained from a COVID-19 dataset that is available on GitHub. This data is then subjected to pre-processing because it contains a lot of null, missing and redundant values. The processed dataset is then passed to the proposed Deep Learning (DL) architecture for selecting only important and effective features. In order to make the process of feature selection in DL more effective, we have implemented the Layer-wise Relevance Propagation (LRP) and the Extra Tree technique on each layer of the DL model. The LRP and Extra tree evaluate the importance of each feature at each DL layer and finally, the output of two is combined to get the final feature set. After this, data is divided into training and testing sets in the proportion of 80:20. To introduce the concept of novelty, we have divided the training dataset into three agents (data subsets) which are then passed to three base models i.e., Support Vector Machine (SVM), Logistic Regression (LR) and K-Nearest Neighbor (KNN) separately. This was not the case in traditional models wherein the entire training dataset was passed to classifiers for training purposes.Results: The outputs generated by three base models are then combined by using an ensemble learning voting mechanism to make the final prediction which determines whether a patient is COVID-19 positive or not. The efficacy of the proposed approach is validated using Python software, wherein, it outperforms traditional LR, SVM and Bernoulli NB models by attaining an accuracy of 97.5%.
Conclusion:The proposed multi-agent-based model for COVID-19 detection shows promising results in terms of accuracy and efficiency. By employing advanced techniques such as deep learning, feature selection, and ensemble learning, this approach addresses key challenges in disease detection and offers a significant improvement over traditional methods.