Abstract
Over the last decade and a half, online advertising on social networking sites have received considerable media interest. Data are being posted to these social networking sites every day. The highly dynamic behavior of users in relation to these services is therefore very important to study. In Facebook posts, user comments play a significant role in making decisions about which service or commodity are worth time and money. Due to many user comments being uploaded to these social networking services every day, and the growing value of these comments. This paper aims to analyze and predict user comment volume generated on Facebook prior to publication on the Facebook platform. We model the feedback from users and estimate how many responses a post will get over the next hour. We established a model prediction using the feature selection algorithm and the random forest model. In this situation, we consider the comments from short textual messages that refer to the main topic of the post. Our predictive model was used on numerous data sets, and the following parameters were measured: correlation coefficient, mean absolute error, root mean squared error, relative absolute error, and root relative squared error estimation measurements. In mean absolute error criteria, our proposed methodology was more successful than the existing prediction models for Facebook comments with 24.40% rate.