A framework for feature selection and classification for high dimensional datasets using meta heuristic Swarm Search

Abstract

Classification is considered to be an imperative task in data mining that aims to classify newlineeach instance into multiple sub-groups. The key factor that influences performance of newlinealgorithm is the feature space . Datasets with enormous set of features are exceptionally newlinecommon especially in the area of bioinformatics, complex technological processes, newlineforecasting and object recognition. newlineUsually, enormous amount of features are introduced in dataset, which may comprise of newlineirrelevant/redundant features that are not valuable for the task of classification. The newlinepresence of these features obscures the vital information provided by significant features, newlinethereby degrading the quality of whole feature set. Moreover, as the dimensionality newlineincreases, it also increases space requirement and time requirement to process the data, newlinethereby leading to computational overhead, overfitting and difficulties in model newlineinterpretability. newline