Feature Selection using Mutual Information for High-dimensional Data Sets

To reduce the dimensionality of dataset, redundant and irrelevant features need to be segregated from multidimensional dataset. To remove these features, one of the feature selection techniques needs to be used. Here, a feature selection technique to remove irrelevant features has been used. Correlation measures based on the concept of mutual information has been adopted to calculate the degree of association between features. In this paper authors are proposing a new algorithm to segregate features from high dimensional data by visualizing relevant features in the form of graph as a dataset.