Authors : Mersha Nigus Alemayehu and Doreswamy
Abstract: Feature selection will become crucial, specifically in facts units with a huge variety of variables and features. It’s going to cast off irrelevant variables and boom classification accuracy and overall performance. With the intention to decrease the model’s computational cost and growth its efficiency, it is a good concept to reduce the variety of input variables. This study employs a wrapper approach to discover a subset of features most relevant to the classification problem. Sequential backward series, sequential forward choice and recursive feature exclusion are the 3 forms of feature selection that Wrapper procedures help. Machine learning classifiers inclusive of k-Nearest Neighbor, Logistic Regression, support vector machine and random forest are used to determine the classification accuracy of selected attributes. The findings reveal that the random forest classifier is the excellent and sequential backward selection with seven attributes is the great filtering approach with 99.97% accuracy and a 100% ROC. Finally, the experiment result of the paper inform to government, policy makers and humanitarian organizations to take an emergency action to fix the problems of household who are food insecure and needs emergency action to survive their lives.
Mersha Nigus Alemayehu and Doreswamy , 2021. Feature Selection for Household Insecurity Classification: Wrapper Approach. Asian Journal of Information Technology, 20: 146-151.