Improving the performance of professional blogger's classification

Abstract

Blogging is a useful way for writing online articles and the individuals who are involved in this activity, are called bloggers. A blogger may have many features such as educational background, cultural background, topical interests and can be classified into classes using these features. There can be many factors (affecting features) by which bloggers opt this profession. Classification of professional bloggers and identification of such influential factors is the topic of interest of this paper. We have used Artificial Neural Network for binary classification problem of a bloggers dataset. The Predictive Apriori association rule mining algorithm is used for factor identification. In this paper, results of Artificial Neural Network are compared with the RandomForest algorithm and Nearest-Neighbor algorithm. It is found that Artificial Neural Network outperforms RandomForest (RF) and Nearest Neighbour (IB1) algorithms with the 87% accuracy and 86.9% F-measure respectively. The results of factor identification are compared with the results of Alternate Decision Tree (ADTree) algorithm. It is observed that both ADTree and Predictive Apriori algorithm produced same results for predictive performance measures.