| The rapid development of the research and application of information technology has been promoting the data mining technology research. At present Many theoretical achievements are obtained on the rearch of data preprocessing ,and lots of excellent data preprocessing tools and the corresponding pretreatment algorithms emerge in large numbers.In the pratical applications,the diversity of data mining tools ,the complexity of pretreatment process,and numerous preprocessing algorithms often make users feel confused, the growing selection of data preprocess technologys requires miners grasp of the data mining knowledge.All these give users a burden, and get to different mining results.To solve these problems, intelligent data preprocessing is researched in this paper, Intelligentize mainly embodied in two aspects: intelligent preprocessing and the intelligence of algorithm selection. The main content and innovation points are as follows:1,Theory on different data preprocessing technology is studied and the characteristics of concrete pretreatment process and the possible recommendation forms are analyzed.2,A multi-agent data preprocessing framework is constructed, It's the solution of the poor extension,poor universality and the inadequate resource scheduling problem. This system integrates such as the preprocessing algorithms,user interaction and scheduling;besides,the algorithm recommendation subsystem is designed to give the best algorithm.3,To solve the Intelligentize of preprocessing,we introduce the multi-agent technology, puts forward a knowledge discovery model based on multi-agent.Every Agent is a module of the preprocessing system,and the whole process of pre-processing system is a multi-agent.4,The automation recommend subsystem is added to the process of discretization to avoid the trail and error attempts.5,Design and realize a prototype system based on the multi-agent architecture, and the intelligent recommend module during discretization is also finished. |
PDF Full Text Request