Research On Rumor Detection Mechanism Based On Transfer Learning

With the rapid development of Internet technology,online social networking has become a new service platform for information publishing and sharing,and has gradually become an important part of people's daily life.Because these platforms have low barriers to information dissemination,and information diversification and expression liberalization,it is an excellent platform for the emergence and spread of rumors.Malicious rumors often have certain harmfulness and are likely to cause social panic,which is especially unfavorable to the progress of society and the development of the country.In social networks,the flood of spam,especially the proliferation of rumors,has become an increasingly prominent social issue.The existing rumor detection methods mainly use traditional machine learning methods.These methods require a large amount of data annotation.However,the content of microblog is short and random.These characteristics make large and effective labeled data difficult to obtain,and labeling samples is time consuming and laborious.Moreover,human subjective factors tend to lead to misjudgment of rumors.At the same time,microblog information is updated quickly,and the data that can be effectively labeled is not enough to train a reliable classification model.Transfer learning is a new machine learning method that uses existing knowledge to solve different but related domain problems.The data and methods of spam detection have certain similarities with microblog rumors.With the emergence of transfer learning,the two basic assumptions in traditional machine learning have been broken,providing the possibility to solve the above problems.At present,transfer learning is mainly applied to fields such as image processing.How to effectively implement the application of transfer learning in the field of rumor detection has become a major difficulty.In view of the above problems,this topic identifies and detects the rumors flooded in Twitter,uses the transfer learning method to transfer the knowledge in the deceptive opinion to achieve effective classification of the target domain(the rumor detection field).The main work is as follows:First,data processing and analysis of source and target domain data is required.Secondly,this paper proposes a deep migration model based on Convolution NeuralNetwork(CNN)to detect rumors in Twitter.Specifically,this paper proposes a learning rate adaptive update method to solve the negative transfer phenomenon in the transfer process.Finally,the deep learning framework TensorFlow is used to implement the proposed model,and a more accurate rumor checking mechanism is constructed under the condition that the size of the rumor has been insufficient.After experimental verification,the proposed model achieves more accurate recognition results in rumor detection and can be used as an effective rumor detection method in practical work.