Research Of Data Mining Based On Outliers And Application In Decision-Making System Of Taxation

The thesis did some research on how to build an analytical system of taxpayer activities based on outliers mining on the basis of being familiar to the taxation business, the present taxation information system and technological study of outliers mining. First, the paper expounded the development of outliers mining, pointed out its shortcomings and made some improvements; Next, the paper designed and realized a complete course of its data preparation; Finally, the improved algorithm was applied to the system and had a good effect so as to complete our work of "The system of data warehouse and data mining in State for Taxation of Taian City" .Some valuable works in the paper include:First, based on the algorithm of based-cell detection, the paper presented a dynamic adjustment function on "M" which can work out a problem of outliers fault-making on boundary cell.Second, In order to help users to solve the question better, we put forward a new way of range of "D" in the origin algorithm.Third, on the basis of the above two works, the paper improved on the origin algorithm and had a better result.Fourth, we found a new way to update the data warehouse based on inserting timestamp into the database. Many experiments have been done to achieve this new way.Besides, the paper had used different taxation-data to verify and compare the availability of the new algorithm. The origin algorithm has the same time complexity with the improved one, but the latter achieved a notability in dealing with the problem of outliers fault-making on boundary cell.