Improving User Relationship Prediction in Twitter Metadata Using Aggregate Classification

In today's world, social networks that have become a part of people's daily life, including Twitter, Telegram, Instagram, etc., are increasing and expanding day by day. Therefore, the number of their users is also increasing and as a result, a large amount of data is being exchanged and stored in these network; and this huge amount of data has turned social networks, especially Twitter, into big data. It is very important to manage, organize and prune these big data, as well as to predict the behavior of social network users.
One of the most important and effective methods for predicting user relationships in social networks is classification techniques, which in most of the applications and researches in the background of the research, are still based on criteria such as ‘accuracy; and accuracy of prediction. have weakness In this article, in order to predict the user relationship in Twitter social networks, the cumulative classification method based on voting, which has two basic steps, has been used. In the first step, by using basic classification algorithms including nearest neighbor, decision tree, random forest and simple Bayesian, the outputs of each classification are obtained. In the second step, the final output of cumulative classification is calculated using the voting method. The results of the experiments on the dataset of the Twitter social network and based on the criteria of accuracy, correctness and coverage, argue that the proposed cumulative classification method based on voting has more favorable results than It has other algorithms