Cohesive Feature Selection for Classification and Retrieval of Data

Feature selection plays a vital role in any problem related to pattern recognition. Various tools and methods are used for feature definition, feature analysis, feature classification, and feature cohesiveness. Much of the work has already been done in all these fields earlier baring a few. Feature selection and classification has undergone many challenges from last one decade but still no robust method has been discovered, which can create a nexus between the human perception and the machine intelligence. The basic and fundamental problem that lies with the semantic gap between two agents namely, human and machine is 'perception' of similar kind of information. In the present paper, an attempt has been made to understand the concept of perception and how to replicate the same in the machine in terms of 'Cohesive feature'. Selection and classification of feature has been undertaken with special reference to conditional and class conditional representations and independence. Experimental Results performed on features have shown promising results.