Research On Micro-blog Public Opinion Prediction Model Based On Machine Learning

With the rapid development of the information society and the deepening of the web 2.0 era,the dissemination mode of information has undergone earth-shaking changes,and all kinds of public opinions and rumors have also filled the cyberspace.As an important social media information publishing platform,Micro-blog has long been an important part of the lives of modern Internet users.It is precisely because the content of Micro-blog has the characteristics of fast propagation speed,wide dissemination range,and strong randomness.As a result,the information quality on Micro-blog is uneven,and it is easy to make public information spread in a large amount and widely in a short period of time.Caused the occurrence of public opinion events and affected the cyberspace environment and social stability.Therefore,it is necessary to study the propagation prediction model of Micro-blog's changing public opinion,not only to discover hot spots in Micro-blog in a timely manner,but also to provide necessary help for adopting appropriate public opinion guidance measures.This article uses information from the Sina Micro-blog platform as a sample to make prediction and research.Firstly,the self-made web crawler completes the crawling of micro-blog text data,and performs corresponding data cleaning,denoising and other preprocessing;then,according to the characteristics of micro-blog text,corresponding hot topic statistical methods are designed to obtain micro-blog for a period of time.The content of the hot topic,extracting the data content such as the number of responses and the number of forwarding data,constitutes the time series experimental data used in the prediction model;finally constructs the relevant prediction model,completes the corresponding experiment,and analyzes and studies according to the experimental results.Due to the strong non-linear change characteristics of the Micro-blog public opinion transmission process,this paper selects the BP neural network as the basis of the prediction model and can achieve a better fitting effect.However,the BP neural network prediction model has the disadvantages of high sample forgetting rate,slow convergence speed and easy to fall into local extremes,which may affect the prediction model's operating efficiency and the accuracy of the output results.Aiming at the above problems,this paper designs using genetic algorithm to improve the structure of neural network,constructs an optimized micro-blog prediction model,and then optimizes the parameter selection in the neural network model through the improved PSO algorithm to find the optimal parameter settings.This can not only shorten the time-series micro-blog public opinion prediction model operation time,but also improve the prediction efficiency and accuracy.The final experimental results show that the optimized neural network public opinion prediction model can achieve better prediction results.