Identifying Spam Tweets in Social Networks with Combined Approaches of Feature Selection and Deep Learning
Sina Saffarzadeh, Mehdi Salkhordeh Haghighi · 2024
Spam detection in social media is troublesome because of language features. Spammers can bypass filtering methods by changing their behavior and following legitimate accounts. The practical method for detecting spam is to classify posts based on their content using a text classification method based on deep learning. This manuscript aims to identify spam in social networks based on content. This paper presents a three-step method for detecting spam in social networks. The convolutional neural network (CNN) is used for feature extraction in the first step. Several feature selection methods are used in the second stage, with the majority voting for feature selection. In the third stage, the combined learning and classification methods are used. In the proposed method, the combined feature selection is made using the chi-square method, random trees, and recursive elimination method. Examinations on the Twitter dataset show that the proposed method in spam detection has accuracy, sensitivity, and precision of 99.46%, 99.36%, and 99.32%, The proposed method is more accurate in CNN+SVM, and CNN+SVM+LSTM.