Rumor Detection Based on Attention CNN and Time Series of Context Information

This study aims to explore the time series context and sentiment polarity features of rumors’ life cycles, how use them optimize CNN model parameters improve classification effect. The proposed is a convolutional neural network embedded with an attention mechanism information. Firstly, whole cycle rumors divided into 20 groups by algorithm each group texts trained Doc2Vec obtain text vector. Secondly, SVM used group. Lastly, spatial classification. experiment results show that introduced very effective for rumor detection, can greatly reduce number iterations training as well. accuracy, precision, recall F1 are better than latest benchmark model.