The Cuckoo Search Guided By Correlation And Classi Cation Accuracy

1873 Words Mar 31st, 2016 8 Pages
20th International Conference on Knowledge Based and Intelligent Information and Engineering
Cuckoo Search Guided by Correlation and Classification Accuracy for Dimensionality Reduction
Waleed Yamany a a,c,
, E. Emary b , Aboul Ella Hassanien
Fayoum University, Faculty of Computers and Information, Fayoum, Egypt b Cairo University, Faculty of Computers and Information, Cairo, Egypt c Scientific Research Group in Egypt (SRGE) b,c Abstract
Attribute reduction is a challenge in that is has a big e ect on classification accuracy as the classifier training time, complexity and running time increases as a number of attributes used increases. Commonly attributes in a given data set are correlated, noisy and redundant. In this work, a system for attribute reduction was proposed making use of filter-based attribute reduction using the correlation-based attribute reduction. The cuckoo search optimizer was used to search the attribute space of attribute set with minimum correlation among selected attributes. By running the cuckoo search for a predetermined ratio of iterations, the selected attribute sets are reevaluated using a new fitness function based on wrapper-based attribute reduction. The initial agents/solutions for this second stage of optimization are guaranteed to have minor correlation and hence are the candidate for further improvement towards the classification accuracy fitness function. So, the cuckoo search completes the rest of optimization steps with the initial…
Open Document