Research Of The Fake Social Network Media Content Detection

Today,with the highly developed Internet,mobile information flows at a high speed,and the amount of information covered by social network media such as microblog is growing exponentially,and gradually becomes one of the main platforms for people to share their lives,make friends and get information.However,in the process of information consumption,media content has the characteristics of convenient access,fast communication and low cost,which makes it become the main source and medium of fake public opinion flooding in our country,and affects people's daily life.Fake media content detection technology is a technology for mining,analyzing,identifying and filtering massive information in social network,which can discover and process the fake social network media content in time,and prevent people from being misled by such information when they understand things and live consumption.Based on the traditional machine learning methods and deep learning methods,this paper studies the detection of the fake media content in social network,and constructs a classification model with excellent performance.The specific research contents are as follows:1)In this paper,we uses the traditional machine learning methods to build a fake media content classification model based on ensemble learning.By observing the structural characteristics of dataset,collecting the required data.Making the corresponding text cleaning method.Extracting the text content features of the microblog with the help of regular expressions,and the comment emotional tendentiousness features of each microblog event with the artificial expanded Chinese emotional lexicon.Constructing the feature set with the combination of user features and communication features,then building the basic classifiers with the commonly used machine learning methods.Finally,using Stacking integration method to build a strong classifier.The experimental results show that the accuracy is 93.19%,which is superior to other single classification models,and the effectiveness of the integrated model is proved.2)In this paper,we uses deep learning methods to build a fake media content classification model based on ensemble learning.In view of the complexity and timeconsuming of constructing feature set manually,and the influence of user attitude expressed by comments on the detection effect,this paper proposes a fake media content detection method(ECC)which integrates the representation of microblog text and the comments.First of all,we pre-train the models with sending the fixed size index vectors of microblog text to BiGRU_Attention model,and the fixed size index matrix of each microblog event's comments to hierarchical attention network.After the convergence of the two models,the parameters of the model are fixed,and the index vectors of microblog text and comment texts are sent into the corresponding model again for tensor calculation.The output of the last attention layer is taken as the feature representation of the corresponding text.Then,the two representations are spliced,and sent to the small neural network for training and prediction.Experimental results show that the algorithm can effectively detect the fake media content.3)In order to speed up the application of fake social network media content detection in real life,this paper develops a fake media content detection system of the microblog based on Python's Flask framework.The system integrates data preprocessing,model training,model evaluation and other modules with providing the function of fake detection of the related Weibo topics.At the same time,users can upload their own datasets and train their own models on this system.