A new model based on Dynamic Naive Bayes classifier for retweet behavior prediction

At present, the prediction of user retweeting behavior plays a significant role in perceiving the underlying mechanism of information diffusion in the microblogging network. The problem of retweeting behavior prediction is usually considered as a classification task hence, the effort for the classification of the time-series data in the microblogging network has gained spacious attention. It is challenging to improve the classification accuracy of retweeting behavior prediction models, because the applied classifications still lack a tailored method. To address this, we present a retweeting behavior prediction model based on Dynamic Naïve Bayes classifier that is an appropriate classifier dealing with time-series data classification. In fact, we apply a time discretization approach and incorporate user individual, social and temporal features into our proposed prediction model to estimate accurately user retweeting behavior in the microblogging network over time. The real Twitter(X) microblogging dataset is used to evaluate the performance of the proposed model in the user retweeting prediction. The experimental results show that our proposed classifier due to its simple structure can avoid overfitting problem, and outperforms in compared to other baseline methods.