Development of knowledge discovery system and research of business process for the implementation

点击翻译按钮获取中文摘要

In our research project, we have tried to develop the KDD system that can be applied to the entire KDD process, for the purpose that the system will be used in an actual company. We have achieved the following three points.1.Analytical methodsWe had joint research projects with some companies (retail companies like supermarket and department store) on knowledge discovery. Those companies provided us with actual sales transaction data with customer ID. Using those data we developed analytical methods that provide a useful knowledge, by which the companies can implement an effective marketing action. We developed following three methods mainly.(1)Knowledge discovery on store arrangement in a department store using genetic algorithm(2)Knowledge discovery in brand purchasing pattern using network flow estimation(3)Consumer Behavior Analysis by graph mining technique2.Algorithms used in the methodsWe developed efficient algorithms for the above analytical methods. We developed the following three methods mainly.(1)Algorithms required for data preprocessing(2)Extension of the algorithm originally used in gene analysis technology to the analysis of business data(3)Efficient algorithms for approximating a multi-dimensional voxel terrain by a unimodal terrain3.Developing a softwareWe have developed a KDD software named "MUSASHI", which was released on the web of "musashi.sourceforge.jp" on 10 July, 2003. Most of the analytical methods and the algorithms mentioned above were already implemented in MUSASHI.More than 60,000 of people had visited on the web page and the software had been downloaded about 5,200 times since opening the web page, which shows our research result is generally accepted.