The Applying Of Outlier Detecting Technology In Evaluation For Coal Value Added Tax

On the background of national economy’s fast increasing and taxation’s quick expanding, it will effectively strengthen the administration level of coal VAT and ability of against tax evasion by means of using data mining technology to analyze the business data, opening out the potential outlier data object, analyzing and deciding possible abnormal coal trading business. This paper designed a evaluation system of coal VAT administration-ESCVA which based on outlier detecting arithmetic and D-S evidence theory, accomplished the implementation and test of ESCVA’s outlier detecting arithmetic.Considering the practical request and special factor of data-gathering for tax administration, this paper used an improved outlier detector arithmetic which based on similar coefficient sum. The improved arithmetic fully considered the factor of the discrete degree among different attribute influencing to the abnormal degree of data object, solved the outlier-ignoring problem in the old arithmetic causing by simply using arithmetic sum. By using D-S evidence theory’s fusion mechanism, the improved arithmetic did not need to perform outlier detecting by attribute, had higher efficiency. The result of experimentation showed that improved arithmetic of ESCVA had more clearly clustering results.Because ESCVA has multi-data source which come from different system, in the analyzing and designing period, this paper used strategy based on practical situation, richly calculated factors from other system’s upgrading and deploying, divided system into multi levels and modules. This paper discussed the main framework of ESCVA, decided software function of each module in ESCVA’s framework. Data acquisition module achieved data-gathering and standardization for other information system, used strategy of fully using existing software resource for reducing complication of system design and maintenance difficulty. Data Processing module used the improved outlier detector arithmetic for performing abnormal detection in gathered data, and processed the gathered data basing on evaluation method of coal VAT. The design of result representation module obeyed MVC mode, fully considered easiness and readability of interface, brought out the evaluating result for end-user.Proved by tax examination and tax experts group, ESCVA’s outlier detecting arithmetic had relatively higher detecting nicety and applying value.