Methods Of Detect Falsely Making Out Specialized VAT Invoices Behavior Based On Directed Graph

Finding financial transactions is the key point when detecting fake VAT invoices.Capital transaction flow can be abstracted as composed of vertices and edges of the directed graph,so research based on the directed graph of the VAT invoice behavior detection method is very necessary.Firstly,this paper studies the fake VAT invoices behavior representation and detection methods during making fake VAT invoice,the search loop problem of funds from the deal flow will be abstracted into the problem of directed graph loop search with multiple edges.Secondly,it shows the algorithm which is searching the map storage and ring road,and compares much algorithms finding that loop search algorithm that commonly used can not solve the problem with multiple edge search to loop well.Therefore,it proposed the map loop search algorithm based on the depth-first.First of all,associate the capital transaction flow with account information using account numbers with the preprocessed raw data,and find the set of edges with transaction time according to the ascending order.For avoiding repeated traversal of the operation process of all sides,the edge weights will be constrained.Recursive program scheme is used for ensuring traverse multiple effective sides and find all the loops in the graph.The correctness of this algorithm is also verified on a stand-alone computer.But when the sides number increased to more than 1 million,the result can't be obtained because of memory overflow.After this,this paper also proposes a loop directed graph search algorithm based on Spark,it preprocessed the original data with a abstracted GraphX graph data structure in Spark,and then verify the correctness of this algorithm in the local mode with data caching,tail recursion and GraphX framework with Scala language.However,the algorithm of time efficiency was lower than originally proposed algorithm in the same data set.Optimizing loop structure and creating index for improving the efficiency of this algorithm.After verifying the result in local mode and using the number of edges for map data more than 10000 sets in Spark cluster,the result shows that the optimized algorithm can not only solves the multiple scale edge searching problem to loop diagram and has good scalability,but also improves the execution efficiency.At last,this paper designs and implements a friendly interface and strong operability system of the VAT invoice behavior detection.After testing,the system can search out of funds from the transaction flow and capital loop to check out the details of the loop associated with the account,and this both provides an important basis and save the audit cost for the audit department to detect false VAT invoices.