The Research And Implementation On Visual Data Mining Technology

With the rapid development of database technology and widely used of the database management systems, more and more data was accumulated. Data mining technology is an effective way to extract potentially valuable knowledge from these massive amounts of data. After more than ten years of development, a number of classic data mining algorithms have become steady. Besides, data mining standards have been gradually constituted and refined. The data mining technology has begun to be widely used in all trades and professions. In order to make the data mining technology meet diversified user needs, visual data mining techniques have been proposed, which enable different stages of the data mining process to be displayed to users with visualization means and allow users to assess the results of data mining more easily and effectively. Visual data mining technology has become a hot research in this field.In this paper, theoretical knowledge, methodology and techniques related with visual data mining are introduced. In addition, the current major research fields of visual data mining are described in detailed. What's more, the development trends of visual data mining are pointed out. Based on above introduction and analysis, corresponding and feasible design ideas are proposed on data visualization, data mining process visualization and data mining result model visualization respectively. Finally, an open and extensible visual data mining prototype was implemented to verify and assess the research work.This paper completes development and implementation of the main functional modules and common components of the proposed visual data mining prototype with Java, JFreeChart and other mainstream technologies based on the common Netbeans 6.1 platform. The prototype provides basic function modules including data visualization, data mining process visualization and result model visualization. In the implementation process, a mapping component between data mining algorithms and GUI was built with DMX. It can provide the GUI access interface for various data mining algorithms based on the component reuse technology. In addition, the results of data mining models were stored according to the PMML standards and a common PMML parser component was designed to analyze various data mining models. With the help of the PMML parser, the data mining model visualization module can achieve seamless exchange and share of result models and visualization capabilities with other data mining software products and tools. In the end of this paper, an open and standard data set was used to assess the visual data mining prototype. The experiment shows that the prototype can provide users with intuitive, understandable diagrams and help them to get a whole picture before data mining. Furthermore, in the mining process, it can help users participate in the implement of data mining algorithms and finally make effective evaluation of the results of data mining with visual graphs so as to enable various users to get a better understanding of the visual data mining technologies.