부산대학교 도서관

Retweet is a Twitter feature use for information propagation to the other user over the platform. During any political event it extensively used for campaigning, spreading information, show political agendas by political parties. The interesting question arises how to predict the retweet probability of the posted text. For this numerous features have been applied in the past. This work aims to find the optimal set of features for retweet prediction using meta heuristic algorithms. Feature selection is one of the important aspects in text classification to reduce the time complexity of the scalable data. We have used genetic algorithm and particle swarm optimization to obtain the minimal set of features to predict the virality of the tweet. We have used three machine learning models K Nearest Neighbors, Naive Bayes and Support Vector Machines to measure the performance of the meta heuristic algorithms in context of features reduction.