Dimensionality Reduction Using PCA and CUR Algorithm for Data on COVID-19 Tests

Enríquez, Marco; Naranjo, Samantha; Amaro, Isidro R.; Camacho, Franklin

doi:10.1007/978-3-030-68080-0_9

Cited by 2 publications

References 7 publications

(10 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

Coronavirus Disease Predictor: An RNA-Seq based pipeline for dimension reduction and prediction of COVID-19

Iqbal

Kumar

2021

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

SARS CoV-2, the novel coronavirus behind the COVID-19 infection, has caused destruction around the world with human life, detecting a range of complexity which has knocked medical care specialists to investigate new innovative solutions and diagnosis strategies. The soft computing-based approach has assumed a significant role in resolving complex issues, and numerous societies have been shifted to implement and convert these innovations in response to the encounters created by the COVID-19 pandemic. To perform genome-wide association studies using RNA-Seq of COVID-19 and identify gene biomarkers, classification, and prediction using soft computing techniques of Coronavirus disease studies to fight this emergency pandemic in the epidemiological domain, and disease prognosis. The RNA-Seq profiles of both healthy and COVID-19 positive patients’ samples were considered. We have proposed an integrated pipeline from bioinformatics in-silico phase for-omic profile data processing to dimension reduction using various prominent techniques such as formal concept analysis and principal component analysis followed by machine learning phase for prediction of the disease. In this experimental research, we have applied different eminent machine learning techniques to implement an effective integrated model using Classifier Subset Evaluator (CSE) followed by principal component analysis (PCA) for dimension reduction to select the highly significant features and then to do the classification and prediction of Coronavirus disease, different eminent classifiers have been applied on the selected features. In this analysis, the Hoeffding Tree model found the topmost performance classifier with a classification accuracy of 99.21% as well as sensitivity and specificity of 99% and 100% respectively.

show abstract