| With the advent of the Internet age,people are more and more inclined to express their views on the Internet.Campus network behavior is an important part of the national network.In this paper,the main voice of the campus network-micro blog campus network behavior data collection,analysis and prediction of the collected data,to further understand the campus network public opinion,so as to realize the timely grasp of College Students’ public opinion.In the clustering algorithm module,this paper selects the hierarchical clustering algorithm which is less dependent on the initial centroid.Based on the commonly used hierarchical clustering algorithm,this paper proposes and improves the density circle cut method to improve the hierarchical clustering.In this paper,the concept of parameter density is used to divide the data points in the data set into regions,which greatly improves the shortcomings of the hierarchical clustering algorithm.Through the relevant experiments,we can get that the optimized hierarchical clustering algorithm has better accuracy.The campus network behavior prediction in this paper adopts neural network algorithm.In the neural network module,the experiment compares the traditional BP prediction model,RBF prediction model and RBF algorithm model after PCA dimension reduction.Because PCA dimension reduction can greatly simplify the input cost,it is found that the PCA-RBF model proposed in this paper is better than the traditional BP model and RBF model after PC dimension reduction A-rbf model has better prediction effect in the campus network public opinion data prediction and analysis.In this paper,the hot topics of microblog in a certain period of time are taken as the time series data of network public opinion heat prediction,and the ar-rbf combination prediction model is constructed.The optimal historical data is selected as the training samples of RBF neural network and ARMA time series,and the weight of ar-rbf combination prediction model is determined by linear optimization method.The public opinion prediction is carried out by comparing with the traditional ARMA single model and RBF single model Compared with the error,ar-rbf prediction model has lower average error,and can better predict the topic popularity of microblog. |
