No Detail Found!

Performance analysis of decision tree extended classifiers (DTEC) on clinical dataset of covid prediction using optimal feature selection

Author(s)
P.S.S. AKILASHRI Department of Computer Science, National College Affiliated to Bharathidasan University, Tiruchirappalli, India
G NITHYA Department of Computer Science, National College Affiliated to Bharathidasan University, Tiruchirappalli, India
Abstract
In Computer Science Technology, Artificial Intelligence, Data Analytics and Machine Learning plays a predominant role in decision-making strategies. By choosing suitable method and algorithms from any of these fields, a good decision can be taken. In health care industry, at present, prediction of Covid 19, is still a very big challenge, since false positive, false negative metrics are occurring frequently through various covid test. Our main aim, in this proposed work is to increase the prediction of accuracy of Covid, by considering the optimum number of features alone, by taking into consideration the laboratory measurement, and also by clinical test understanding. When target variable is binary, the classification algorithm based on Tree Based Extended Classifiers like Random Forest, AdaBoost, XGBoost can be proposed with necessary features. The results are observed from the proposed algorithms, that gets trained using the training dataset using standard data repository and it is being tested with the testing dataset. By analyzing the performance metrics, the obtained results showed that the prediction accuracy is increased and also false positive and false negative are reduced. In the proposed work, the tree based extended classifiers of Random Forest and Extended Gradient Boosting produces maximum 92% accuracy with 11 features using Gini Index. Apart from accuracy, the metrics such as false positive and false negative are playing the important role. In this proposed work, the false negative is as low as 5 out of 14 by XGB, and false positive with the minimum value of 3 out of 106 using Random Forest. Thus, these methods of covid predictions are useful for health care community, if it is being utilized in an efficient manner.
Publication Details
Page(s) 6465-6481
DOI DOI not available
Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 21, Year: 2022

Request the full text from PASTIC

Keywords
Covid, Feature selection, Confusion Matrix, Tree Based Classifier, Binary Classification, Clinical data
0

Citations

0

Downloads

88

Views