Feature Selection for Embedded Media in the Context of Personification

Now-a-days increase in embedded media applications attracted huge number of users because of no limitations and easy chat conversations. Most of the information is shared between known to unknown persons resulting in safety issues. Mainly embedded media data shared among users is high dimensional consists of comments, text messages, images and association data about users which basically describes context and relationship between embedded media users. In addition scale, noise, errors and incompleteness exacerbates the ever challenging problem for feature selection. As a result it is very much needed to analyse the importance and relevance of features to understand the personification of user identity. This paper proposes a machine learning technique for feature selection of safety domain features in a vector for personification. Detailed illustration of the importance and relevance of attributes towards better understanding of the personification is studied. Proposed approach is tested on real world embedded media data like Whatsapp and BlogCatlog to demonstrate findings. In addition, analysis of the user to user relationships and user to data relationship towards feature selection criteria is done. This proposed solution can improve the performance of feature selection of embedded media data prioritizing personification.