Research On Topic Detection And Tracking Technology Of Network Public Opinion

The Internet has become one of the main channels of people’s daily access to information because of its good interactivity,strong real-time and so on.As a result,The status of the Internet on the guidance of public opinion and the influence of the people consciousness is increasing.Network public opinion has gradually evolved into the core element of social public opinion,and its influence on society can not be ignored.However,Network public opinion is different from the Traditional social public opinion,Network public opinion has burst,the huge amount of data and other miscellaneous features.therefore,It is difficult to capture public opinion information on the network efficiently and efficiently.So,the corresponding network public opinion monitoring system emerges as the times require.The system provides good support for relevant government departments to grasp information and implement efficient and scientific decisions.and then,effectively guiding public opinion and maintaining social stability and harmony.This paper focuses on the research and analysis of topic detection and tracking technology in network public opinion,and improves it on the existing technology means,so as to achieve better application effect.The research work of this paper mainly involves the following aspects:An incremental topic clustering based on JRPCL is proposed for topic detection.This paper studies and analyzes the Classical Sing-Pass algorithm and incremental K-Medoids algorithm Firstly,In the light of text input sequence sensitive and initial center point selection problem,finally selected the improved RPCL algorithm to generate the initial cluster s to avoid Select task of initial center points.At the same time,the Prim algorithm is used to cluster the new data in a certain range,and then the incremental clustering is realized.To some extent,the defects caused by the input sequence sensitivity have been improved.The accuracy and response time of the clustering algorithm are improved effectively.Which makes it suitable for large-scale text clustering.Finally,the experimental comparison shows that the proposed algorithm has good practical application effect.An adaptive topic classification based on dynamic threshold and classifier integration is proposed for topic tracking.an adaptive topic tracking algorithm based on time information and an adaptive topic tracking algorithm based on feedback learning are studied and analyzed Firstly.Aiming at the shortcomings of dynamic threshold setting and classifier ensemble,this paper proposes to add the nearest time to report interval as a factor.Meanwhile,the concept of selective integration is introduced into the ensemble of base classifiers.And then,reducing the false alarm rate and false alarm rate of classification algorithm.Finally,the experimental comparison shows that the proposed algorithm has a better practical application value.On the basis of related technology research,this paper designs and constructs a network public opinion topic detection and tracking system.Expounding the realization and function of each module of the system.Through the actual operation of the system.It is proved that the scheme is highly feasible.The effectiveness of the above method is further verified.