Research On Visualization Of Massive Tax Data

With the rapid development of computer science and technology, information visualization becomes more and more popular and related with people’s daily life.Information visualization, which is an interdisciplinary field, focuses on studying the visual presentation of large-scale non-numerical information resources, providing assistance for understanding and analyzing data for people. So far,a variety of visualization methods have been proposed. And kinds of visualization techniques have been applied to many domains,such as finance, business, literature, abstract concepts, etc. Nevertheless, visualization of massive tax data is still scarce.Tax business management data, generated by the tax system, has features of mass and complexity, and tends to be multi-dimensional. The existing information visualization methods are not well positioned to meet the needs of its visualization. For example, using parallel coordinates,it will produce diagrams which are in chaos,difficult to find the underlying data structures and patterns. There are two major reasons for disorder,one is the non-integrity of tax data, namely that the missing value problem in dataset, the other is that the correlation of two adjacent dimensions is not tight enough. Aiming at those, the paper proposes an incomplete and mixed data oriented method to reduce the confusion degree.Specifically, (1) a missing value filling method based on association rules is put forward to enhance the integrity of tax data; (2) the outlier definition is given to describe the relationship between dimensions, also corresponding calculation algorithm about outlier; (3) an algorithm for the best order of dimensions is presented in order to achieve better visualization of parallel coordinates; (4) a measure on confusion is proposed, and is taken as the metrics judging visualization quality.Finally the paper compares the confusion reducing method with original parallel coordinates by experiment. The experiment shows the method proposed in the paper has better visualization effect.