Research And Implementation Of Visualization For Cluster Process Based On Parallel Coordinates

Visual Data Mining is a data mining technology integrated with visualization, this integration makes the application of data mining more intuitionistic and visual, combine human cognitive ability, creativity and domain knowledge with the data mining process, their advantages can be exploited fully, not only embody the key role of the user, but also improve the efficiency and accuracy of data mining.The main works of visual data mining research are listed as follows: data visualization, result presentation and the visualization of data mining process. At present, works at data visualization and result presentation are plentiful, but the works of process visualization is still not perfect. Research on visualization of data mining process is still stay in a coarse-grained aspect. Users can control data inputting and parameters setting only, while works of fine-grained visualization are not enough, there are lack of dynamic visual display of algorithm steps and the middle result of algorithm.This paper focuses on the visualization of data mining process. Firstly, introduce the concept of visualization and data mining technology, then study deeply on visualization and data mining algorithm, put forward a fine-grained visual data mining model, design and implement an extendible data mining prototype system, finally, using parallel coordinates to analyze the clustering process. The main tasks of this thesis are as follows:(1) Study the general descriptions and basic requirements about the visualization of data mining process, and analyze the significance of the process visualization and presents a visual process model in a fine-grained aspect.(2) Analyze the advantages and disadvantages of existing data mining tools, and characters of data mining process are also taken into consideration, a design scheme of data mining prototype system based on MVC architecture is presented which implemented loosely-coupled and reusable software architecture provides a friendly, flexible, easy reuse and extensible environment of data mining application.(3) Achieve the process visualization of clustering algorithm by parallel coordinates in the prototype system, and show the direct cluster analysis and interactive clustering process by experiment, the results demonstrated that the data mining process through visualization, it makes certain sense to achieve user participate the data mining actively, and improve the efficiency and accuracy of data mining.