Analysis of Feature Selection Algorithms and a Comparative study on Heterogeneous Classifier for High Dimensional Data survey

摘要

This paper focuses on the analysis of various feature selection algorithms and a comparative study on heterogeneous classifier predictive accuracy problems to work with high dimensional data. Especially we conduct experimental comparisons of IBK (KNN), SVM, NBTree and J48 on KDD Cup99 intrusion detection dataset and one cancer disease diagnosis microarray datasets and analysis their performance with vote generalizations. Based on the fact a large number of features can cause a noise of data and degrades a performance of learning algorithm.To tackle these problems identifying a suitable feature selection method is essential for a given machine learning algorithm tasks. So feature selection plays a great role in intrusion detection, bioinformatics, and medical data analysis. Thus this paper deals the application of best feature selection techniques to improve learning algorithm predictive accuracy in microarray dataset and KDD (Knowledge Discovery and Data Mining Tools Conference) Cup 99 dataset with a respective classification and feature selection algorithms. basically, this approach shows the application of feature selection algorithms when a large number of features represented in a small sample data and small numbers of features represented with a high number of samples by taking the above two different datasets.