A PERFORMANCE ANALYSIS OF EVOLUTIONARY BASED RANK FEATURE SELECTION AND PRINCIPLE COMPONENT ANALYSIS
In Data Mining and Machine learning the first step of analysis undergo preprocessing of the features in a given space. Managing large number of features always increase problems for a model in terms of both accuracy and time complexity. Many research have been done on various feature selection , feature construction and feature reduction methods to reduce the number of properties to improve accuracy of the model built. In this paper a performance analysis have been done on Symmetrical Uncertainty (SU) feature selection embedded with evolutionary algorithm and Principle Component Analysis (PCA) for two datasets of Autism Spectrum Disorder from kaggle repository and data collected from special schools and tested the accuracy with specific classifier. The implementation results on selected UCI datasets with binary, continuous variables it explained that principle component analysis works best with expected accuracy. For limited data set especially with categorical variable features alike Autism spectrum disorder data filter selection with evolutionary algorithm formerly proposed by the author  is useful through decreasing the volume of initial features and improved accurate result by implementing better detection performance in the classification methods relating with other feature selectors.