The Research Of Multi-Relational Data Mining Technology And Its Realization In Tax Assessment

As a kind of method and technology, Data Mining (DM) aims at discovering the latent rule from massive amounts of data and extracting the useful knowledge. In recent years, DM has gotten domestic and foreign widespread concerns and has been becoming most hot forefront in the field of information systems and computer science. As traditional DM technology is based on single relational foundation, can not fully meet the application in circumstance of complex data reality, this paper proposes a multi-relational data mining technology.At present, the tax assessment has become an important part of our country tax revenue management. With computer technology and networking applications, our country tax system has realized data centralism model above the provincial level, and formed many application systems sharing a network pattern. The functioning of these systems has had the massive service data, and how to integrate, analyze and mine the data to support current and future revenue scientific management and decision-making will become the important work of the current information-based tax. The tax assessment information system established by traditional technical means has many problems, and the use of MRDM can solve these questions effectively.Firstly, through research on the process mold and the technique mold of data mining, this paper analyzed the application and actuality of tax assessment system. As the assessment index in Tax Assessment can be established and maintained automatically by the computer system thus enables the production of assessment object to have certain scientific nature and notarization, taking this as the aim this paper chooses the right data mining method–sorting. Secondly, after studying various arithmetic of sorting, has ascertained to use the Supervised Learning In Quest (SLIQ)of multi-relational decision tree,but owing to the algorithm can not directly use the data of Database -Management System (DBMS) and the calculation volume in the construction decision tree process is oversized, so chooses to use their improved algorithm QLIQ to implement. Finally, according to data mining's process model and in virtue of the Analysis Manager which is furnished by the extensively used large-scale multi-relation data system, this paper designs the filtering system with tax assessment object basing on QLIQ, and realizes the automatic establishment and maintenance of the tax assessment index and promotes the rationality and justice of the assessment object. The experiment indicates that QLIQ Algorithm can effectively solve the problems existing in TCMIS to meet operational needs, such as the index choice difficultly also needing manual configuration as well as assessment object production lacking scientific basis and so on, and it is also highly efficient and can be easily implemented. Subsequently, at the end of the paper, a subsystem is also briefly introduced that the outcome is to be fed back by the finished tax assessment.