Algorithm Based on Improved Naive Bayesian for Predicting Microblog Behavior

This paper is aimed at predicting microblogs' content and three kinds counts of behaviors that including forward counts, comment counts and like counts, which analyse the overall characteristics of microblog, and propose an algorithm based on improved naive Bayesian-NB for predicting microblog behavior. Calculating microblogs' feature words based on TF*IDF, we classify microblog by LDA algorithm, to find the right category set. Select feature words used frequently from microblog, as predictive properties, and construct an improved naive Bayesian algorithm for predicting. The result of experiment shows that the recall rate, precision rate and F1 evaluation value are improved.