Information theoretic based Interactive algorithms for Feature selection

Abstract

newline Feature selection is one of the important techniques in data newlinepreprocessing for data mining It reduces the number of features removes newlineirrelevant redundant or noisy data and brings the immediate effects for newlineapplications Reducing the dimensionality of the data reduces the size of the newlinehypothesis space increase classification accuracy and reduce the computation newlinetime The selection of a small number of highly predictive features is used to newlineavoid over fitting the training data The high dimensional data often contains newlinemany redundant features Both theoretical analysis and empirical evidence newlineshows that irrelevant and redundant features affect the speed and accuracy of newlinelearning algorithms and thus should be eliminated as well newlineFeature selections have also been applied to small and mediumsized newlinedatasets in order to locate the most informative features for later use newlineMany feature selection methods have been developed and are reviewed newlinecritically in this thesis with particular emphasis on their current limitations newlineThe leading methods in this field are presented in a consistent algorithmic newlineFramework newlineThe choice of selecting particular classification method is based on newlinethe classification accuracy computation time robustness scalability and newlineinterpretability newline