Improved Blog Classification Using Multi Stage Dimensionality Reduction Technique

Blogging is a community-based effort and blogs carry rich information on a specific context. It serves as a platform for individuals to voice their knowledge, skills, thoughts, and feelings which become ingrained in our day-to-day lives. Improved blog classification using multi stage dimensionality reduction technique is proposed. A blog can be easily classified using the tags provided by the blogger. The blog contents can also deviate to a related topic as it can be written by novice content writers also. In the proposed method the blog contents are represented as a collection of text features. The blog representation and preprocessing stage transforms the blog contents into feature matrix. In the feature reduction stage the vital features which are pivotal in its class identification are recognized using term weighting technique and information gain. The pattern reduction stage uses leader algorithm to select only marker data from the data clusters and minimizes the number of patterns from the large amount of blogs. The reduced patterns are fed to compact pattern reduction and classification stage where a novel pruning algorithm N2PS is used to optimize the classifier by reducing the features of dataset. The multi stage dimensionality reduction technique improves the speed of the classifier. The proposed method is implemented and validated on a live dataset. The results outperform the existing methods in terms of accuracy and training time.