Research And Implement Of The WEB-Based Analysis Platform And The Measurement Model On Visualization Data Mining

"Big data" has recently become a very important concern in computer information technology industry. How to use the existing technologies to dig out the abundant information hidden in the data, and create higher economic value, has become the key of "big data". Visual data mining technologies represent data onto images and graphics, and make common user having little professional knowledge with data mining can also participate in mining process, and interactive operation interfaces make the process more convenient. Variety visualization technologies are different from each other in implementation principle, forms, applicable object. How to scientifically evaluate visualization technologies has become the key to promote the research and development of information visualization.By analyzing the Tree-map visualization’s characteristics from the basic idea, layout algorithms, interaction design, visual performance and application scope, this paper established a measurement model basing on the Analytic Hierarchy Process method, and built a test console basing on WEB, preliminarily evaluated the Tree-map visualization. The main research work focused on the following three points:1. Basing on the summary about the graph characters, advantages and shortages of Tree-map, this paper extracted eight metrics from three aspects:scale of dataset, user cognitive complexity and visual performance, computed the weights of all metrics by AHP, and established a measurement model about Tree-map visualization.2. According to the measurement model, this paper designed and implemented an evaluation system of Tree-map visualization which consists of three modules: data preprocessing, Tree-map visualization and measurement model. JSON and XML files are supported and four Tree-map algorithms were implemented:Slice-and-dice, Squarified, Strip and Spiral algorithms. Also the measurement model was implemented.3. The experimental parts visualized several datasets of different types by four Tree-map layout algorithms, and measured the visualization results of each datasets respectively. Based on the horizontal and vertical comparison of the measurement results, this paper verified the reasonability and validity of the measurement model. It turned out that the measurement model could be used to evaluate the advantages and disadvantages of different Tree-map layout algorithms, and then helps to select the optimal forms for different datasets.