Visualization Technology Research On Timing Information Of Network Hot Events

In the era of big data, the occurrence, development, climax and extinction of hot events are more and more rapid, the factors that influencing the development of events tend to complicated, events involving information become more diverse. In face of massive,complex and fast updating multimedia, how to effectively eliminate redundancy, grasp the development trajectories and influencing factors, has become important issues that need to be solved for understanding, guiding and controlling hot events.In this paper, we chose online hot events as study category, analyzing web pages, images,video and other media involved in hot events. At first,this paper analyzes and proposes involving methods and timing multimedia visual models, on the basis of above work, this paper mainly researches the text model, feature selection, text classification and other key algorithms and relizes. The main contribution of this paper is reflected in the following aspects:(1)We propose timing multimedia visualization model of online hot events.Based on the analysis around a variety of timing multimedia features of hot events, proposes online hotspot multimedia visual model from points of views as coving the timing, relevance, senior statistical characteristics, spatial characteristics of geographic perspective. The model with integrated multimedia stratiform Theme River,composition tag cloud and spatial situational diagram core content, regards the subjects involved in the incident as the main body, relizing the real-time performance and evolution tracking.(2)We propose a feature selection method based on improved TF-IDF algorithm.We use SVM classification method, using the training set to build an effective classifier, divide the news into a variety of established types.To improve classification accuracy and efficiency,we improved TF-IDF feature selection method,according to the web news “ inverted pyramid” structure,the news text is divided into two parts,namely, the proportion of different weights defined and used in the classification. Improved methods can calculate feature items weight according to different parts of the news features information entropy, thereby increase the accuracy of text classification.(3) We propose a method for named entity recognition based on a rule of word frequency statistics.Combined with the characteristics of the coverge of online hotspot events,we proposes a method for named entity recognition based on a rule of word frequency statistics, which have some effect for specific information extraction, with fast speed, providing data support for space situational presentions of the behavior of the main subjects of hot events.This paper finally construct the timing sequence of multimedia hotspot visualization system, which integtating the above algorithm to achiwve a visual environment that can represent a variety of timing integrated multimedia information.We validate the validity of the proposed visualization model, provide a foundation for the study of the application.