The Design of the Network Public Opinion Monitoring System based on Cluster Analysis

Aim: To design a network public opinion monitoring system combined with Chinese information processing and to achieve text cluster and theme discovery by introducing an improved K-means approach for solving the problem that exists in the study of network public opinion mining. Methods: An improved K-means approach is set up by constructing one network popular feelings Supervisory system based on the cluster analysis, and introducing key technologies in detail, and the key link in K-means algorithm (the cluster starting value's choice and the isolated point rejection) has been improved. Results: The system can monitor network public opinions by gathering precisely network resources like webpages, forums, blogs and news commentaries and combining Chinese information processing technology such as webpage purification, Chinese word segmentation, establish of vector model, feature selection, dimensionality reduction and text clustering. The general idea of the improved algorithm is that the improved algorithm can automatically compute the clustering results k after users input the initial number k of the cluster and maximum value K-max. Conclusion: Based on the cluster analysis, a network public opinion monitoring system is designed and an improved K-means approach is introduced. The next key step of network public opinion mining and the integration of the key technology is to implement concretely the algorithm and create a set of automatic system of network public opinion gathering, analysis, monitoring and early warning.