Social networks today are great source of data which can be used and analyzed in different ways. In our project the main goal is to predict the behavior of the users, more accurately said: we try to predict what will a particular user tweet in the future, when given a training set of existing URL mentions. The topic has been previously explored in the paper Outtweeting the Twitterers-Predicting Information Cascades in Microblogs by Wojciech Galuba, Karl Aberer, Dipanjan Chakraborty, Zoran Despotovic, Wolfgang Kellerer and our work is the extension of their research. As mentioned in their paper, accurate predictions can be used for many purposes. Some of them are: personalized recommendation, ranking and filtering of the stream of tweets that the user sees etc. Also. it can be used in predicting in virality of a certain URL which can help in marketing campaigns for predicting the success of that campaign. One of the main tasks on this project is: Build a framework which can rank, evaluate different models of predictions with online data. We made a tool/framework in Java that can work with different models of predictions, the evaluator just need to specify the model of prediction with very little code and the framework offers a lot of support for calculating usually used parameters in the model so the prediction is made really easier once the model is known. However in order to test if our framework works correctly, we implemented two models which will be described further in the report and we were working with offline data.