Real-time monitoring for university network public opinion information based on improved deep learning

To improve the coverage of public opinion concepts and the accuracy of monitoring public opinion popularity, this paper designs a real time monitoring method for university network public opinion information based on improved deep learning. Normalise the online public opinion data of universities and use an improved stacked denoising autoencoder to extract features from public opinion information. By introducing sparsity constraints, the model avoids learning redundant features and improves the generalisation ability of feature representation. Based on the SEIR model and propagation threshold calculation, a dynamic monitoring model for online public opinion in universities was constructed to predict the development trend of public opinion in real time and effectively warn of the malignant evolution of public opinion. The experimental results show that the concept coverage of this method can reach a maximum of 0.991, and the accuracy of public opinion heat monitoring can reach a maximum of 0.95.