A hybrid Genetic K-Means Algorithm forFeatures Selection toClassify Medical Datasets

Relevant features selection is become primary preprocessing step for buildingalmost intelligence machine learning systems. Feature Selection (FS) is more andmore important in many applications such as patterns recognition, medicaltechnologies, data mining environments and others. The main objective of FS is tochoice the important features among multi set in order to building effective machinelearning models such as pattern analysis model by cancelling irrelevant or redundantattributes. An addition to that, there is a fact that the efficiency of the desired systemis very sensitive to choose of the features that effect on classification or any analysisprocedure of small or high dimensional datasets. Furthermore, the analysis of medicaldatasets has become growing claiming problem, due to huge datasets that cause timeconsuming and uses additional computational effort, which may not be suitable formany applications.This work attempts to introduce a hybrid genetic k-means of feature selectionalgorithm for multi medical diseases datasets. The proposed algorithm uses a geneticalgorithm combine with k-means algorithm as a powerful tool to select the relevantfeatures from different large medical datasets of Mirjan hospital diabetes, heart andbreast cancer diseases which play the important role in maximum the classificationaccuracy and efficiency of the system. Experimental results show the efficiency ofthe proposed system for the used datasets and satisfy maximum classificationaccuracy performance compared with others states.