Research On Date Visualization System Based On SOM Clustering Algorithm

Cloud computing and big data is one of the hot issues of the present study. it willproduce large amounts of unstructured data for every moment today. However, processingand analysis of these data is still facing many challenges.on the one hand, the huge amountof data, various types of data,a big part of unstructured and semi-structured data,the dataprocessing is a challenging job. On the other hand, there exists a strong timeliness of thesedata, because it generates new data all the time constantly, which requires high speed dataprocessing, otherwise the data treatment will be not the slightest significance.Visualizationtechnology by the large amounts of data with graphics intuitive display.Through theobservation of visual graphics, it can be found that large amounts of hidden data andexcavation of the secret, Provides reliable technical support for auxiliary decision-making,and can bring enormous social and economic benefits to the community and businesses.This paper describes the necessity and significance of research on visualizationtechniques, and characteristics and development prospects visualization techniques; Thenintroduces the concept of visualization and classification, focusing on text visualizationtechniques introduced. In addition, it makes a detailed presentation of the traditional SOM(Self-Organizing Mapping) to analyze the shortcomings of its existence, and based onwhich winning strategies and initialize the weights of SOM clustering is designed toimprove the performance.For SOM clustering algorithm caused by random initialization ofweights clustering to improve the issue time is too long,through analysis of the initialweight calculation, find out the suitable initial weights, can save a lot of time.For SOM clustering decisive strategy did some improvement, solve some neurons can’t win, someneurons won too many problems.Based on the improved algorithm to do a comparativeanalysis of the mass data processing, the improved algorithm has more accurate thanbefore.Finally, the design and implementation of a SOM-based visualization system isdetailed described, WeChat public account as example,based on the characteristics ofWeChat public account information access,clustering results are figured out using theproposed SOM algorithms in the visualization module of jfreeChart,can accurate reflectionof clustering results.