Predicting the Volume of Feedbacks for Blogs using Machine Learning Techniques
Nowadays, the world is evolving by the technological revolution, this can be observed by the tremendous efforts of scientists and researchers who try to solve problems by relying on technical solutions. Social media became the most important channel of communication, also, an essential source of information that plays a key role in people's lives these days. The focus resides on the analysis of these social networks. The huge amounts of available data along with the ever-changing information on these sites make it impossible to perform this analysis manually. This study focus on analyzing blogs based on a number of features to predict whether a blog receives feedback or not. In addition, it proposes applying different feature selection techniques along with different machine learning algorithms for the classification phase utilizing various classification methods (i.e. Naive Bayes (NB), Support Vector Machine (SVM), Neural Network (NN), K-Nearest Neighbor (kNN), AdaBoost, Bagging, Random Forest (RF), and J48 Decision Tree). Furthermore, the experiments test the process of discretization to categorize the blogs based on their comment volume, as well as, identifying the most influential features that contributed the most to our results. The proposed approach is evaluated using the Blog Feedback benchmark dataset. Experimental results demonstrate that the proposed approach can effectively enhance the prediction accuracy of the blogs feedback volume, while significantly reduce the number of features.
Keywords: Blog Feedback Prediction, Classification, Feature Extraction, Social Media