MIME: Mutual Information Minimizer for Selection of Categorical Features
Sayantan Chattopadhyay · 2021 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT) · 2021
MIME is a feature selection method that can be used to reduce the feature set before clustering of categorical features. The algorithm uses Mutual Information as a decision variable to subset the feature set. This feature set can then be used as an input to KModes clustering algorithm. Since this is a filter based algorithm, it is faster than wrapper or embedded algorithms. The primary objective of this algorithm is to reduce the computation time taken to cluster datasets consisting of only categorical features.