Micro-blog Retweeting Prediction Based on Combined-Features and Random Forest

Aiming at the issue of incomplete features selection in micro-blog retweeting prediction researches, a micro-blog retweeting prediction method based on comprehensive features and random forest was proposed. Firstly, macroscopically, we analyzed the distinguishing features between high-retweeted micro-blog and low-retweeted micro-blog, between high-retweeted users and low-retweeted users, between high-retweeting users and low-retweeting users, and extracted nineteen features as the characteristics of micro-blog retweeting prediction. From the micro perspective, we extracted three local features which included the user's activity, the user's interest to the followee, the user's retweeting interest to the micro-blog, as microscopic features to predict micro-blog retweeting. And then we defined the computed methods of the features. Finally, combined with the features, a micro-blog retweeting prediction method based on Random Forest was proposed. The experimental results on Sina micro-blog datasets showed that the proposed method was superior to the classical classification prediction algorithms such as Logistic, BayesNet and Support Vector Machine (SMO), and had good stability on datasets of different scales.