Research On Public Opinion Classification Technology Based On Heterogeneous Data And Neural Network

With the development of the network,the development trend of public opinion data is showing explosive growth.The network is glutted with huge quantity of pictures,voice,text and other types of data information,and these increasingly complex network data are combined to form a complex data structure to express data information.In public opinion data,it is getting more and more difficult to express data information completely through a single type of data(pictures,text,voice,etc.).In recent years,confronting various data information,neural network has made great progress in the applications related to various data classifications and object recognitions by extracting data information from low-level feature space to high-level feature space through its unique hierarchical structure.The wide applicability of neural networks in various data and research fields,provides strong support for the information fusion in this paper.For a network public opinion information containing multiple types of data,this paper proposes a new public opinion classification model,which uses the neural network model to learn the data features of different types of information separately,and classifies their features after fusion.Through this method,data information can be better classified.In the experiment,this paper respectively uses LSTM and CNN neural networks to extract the features of text and image data,and then classify them after feature fusion.The results show that,the classification of network public opinion data information can be better realized by fusing and classifying multiple types of data features,thus improving the accuracy of classification of public opinion information.On the basis of predecessors,the improvements proposed in this paper mainly include the following three points:(1)According to the analysis of the features of different types of data,different neural network models are used to construct feature extraction models of various types of data.(2)In the face of the trained feature extraction model,the first consideration is to extract the various types of data features from different models into the same feature space for feature fusion.(3)In order to further improve the performance of the classifier,this paper uses weight to fuse different data features.