الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 189 |  |  |
| | | from | 189 | | |
Abstract
Feature selection is considered as a fundamental phase that has to be passed to facilitate cancer classification problems. The main approaches of feature selection are filter methods, wrapper methods, and embedded approaches. Two feature selection frameworks are proposed to help in selecting relevant features and ignoring irrelevant ones. The selected features will be exploited in a classification process. A proposed method integrate k-nearest neighbor (KNN) technique and memetic algorithm (MA), which is dubbed K-nearest neighbor – memetic algorithm (KNN-MA), is applied. Secondly, a wrapper method comprises a forward greedy search, which is dubbed wrapper naïve-greedy search (WNGS) is developed. WNGS method integrates Naïve Bayes classifier (NB) as a learning schema with a forward greedy search method. The two integration strategies are applied to two different breast cancer datasets to validate the performance of the two proposed frameworks. from experiment outputs, the two frameworks achieved high classification accuracies and also achieved tremendous success in minifying features space. from extracted results, WNGS framework overcame KNN-MA framework regarding running time. Both techniques achieved close results regarding selected features and classification accuracy.