A taxpayer tax evasion suspect group detection method based on multi-stage mapreduce model

Abstract

The invention discloses a taxpayer tax evasion suspect group detection method based on a multi-stage MapReduce model, comprising the following steps: firstly, a method for constructing a taxpayer interest-related network based on a coloring graph is proposed by using multi-network fusion; secondly, A reduction method of the taxpayer's interest-related network based on strongly connected components is proposed; then, the concept of antecedent network of interest is introduced, and all extremely weakly connected subgraphs are found in the network of antecedents of interest; finally, a multi-stage MapReduce model is used , build a pattern tree for each maximally weakly connected subgraph and its corresponding transaction edges, then traverse the pattern tree to generate a pattern library, and then perform pairwise matching on the patterns in the pattern library to find all pattern pairs that meet the matching principle, Finally, all taxpayers suspected of tax evasion groups are generated. Among them, the multi-stage MapReduce model can greatly improve the efficiency of the national tax department's analysis of suspected tax evasion due to the use of distributed computing, and recover the loss of tax loss for the country.

Description

Technical field[0001]The present invention relates to a taxpayer's detection method of taxpayers to speculate based on multi-stage MapReduce models.Background technique[0002]With the continuous development of information technology, national tax information has initially established unified electronic declaration, tax payment, approval of integrated technology support and service platform, and generate massive tax data. However, the analysis of the tax evasion taxation tax for enterprises is artificial analysis, and the tax data is cross-regional heterogeneous storage, which is difficult to discover the implicit interest association between taxpayers.[0003]Patent "Based on the tax-based tax-based network model" (CN103383767B) Modeling with coloring weighting maps, and gives five polymerization operators based on edge and node color, combined with these polymer operators found The taxpayer's interests are associated with the smallest network, while simplifying the taxpayer's interest...

AI Technical Summary

Problems solved by technology

[0005] The above existing taxpayer tax evasion suspect group detection methods have the following problems: 1. The constructed taxpayer interest-related network does not consider the problem of high graph processing time and space complexity caused by large-scale nodes and complex edge relationships; 2. .There is a pattern combination explosion problem in graph reduction and graph mining; 3. Patent CN104103011B is difficult to find non-motif structures (such as pentagons, hexagons, etc.)

The above three points lead to their inefficiency when dealing with large-scale graph data

Images