Effective Detection and Forecasting of COVID19 Outbreak Using Optimal Feature Selection Based Deep Network Model
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The most severe disease that has recently spread quickly around the globe is COVID-
newline19. To treats patients at an initial period and save lives, accurate Covid-19 identification
newlineis one of the most important needs in the medical industry. However, addressing the
newlinechallenges in existing deep learning approaches for COVID-19 diagnosis established
newlineon both text and image datasets is crucial for improving diagnostic accuracy and sys-
newlinetem performance. In the realm of text-based models, one major challenge lies in the
newlineinterpretation of unstructured and diverse clinical data, which can lead to variations in
newlineprediction outcomes. Additionally, the scarcity of labeled datasets hampers the training
newlineof deep learning models, limiting their generalization to different patient populations.
newlineOn the image side, challenges include the need for robust feature extraction methods to
newlinecapture subtle patterns indicative of COVID-19 in medical images. Furthermore, issues
newlinerelated to data quality, such as variations in imaging techniques and patient position-
newlineing, can affect the methods capability to provide correct diagnoses. Overcoming these
newlinechallenges is essential for advancing the reliability and effectiveness of deep learning
newline(DL) methods in the diagnosis as well as prediction of COVID-19 outcomes. To avoid
newlinethese difficulties and recover the system presentation, an efficient detection model is
newlineessential. Firstly, this thesis analyses how COVID-19 spreads globally and suggests a DL
newlineapproach based on artificial intelligence (AI) to identify COVID19 patient instances,
newlinefatality rates, and recovery rates from the disease. Pre-processing techniques are ini-
newlinetially used to eliminate any undesirable data, such as prepositions, links, hashtags, etc.
newlineThe characteristics are then removed from the pre-processed dataset using the term frequency inverse-term frequency (TF-IDF) and bag of words (BoW) approaches.