Financial Fraud Community Detection Algorithm Based On Probabilistic Model

As financial crimes such as telecommunications fraud,illegal financing,and drug smuggling and trafficking become more and more rampant,money laundering has developed from a means of fund transfer to an independent financial crime,and it has become more and more intense,causing tremendous damage to various countries and organizations around the world.Therefore,it is more and more necessary to build a more efficient and accurate anti-money laundering system.Traditional financial anti-fraud systems,including anti-money laundering and risk control,are basically based on expert experience.The various characteristics of financial fraud are summarized manually,and then the detection and screening are carried out one by one.The methods are more diversified.It is particularly difficult when the technology is more advanced and the crime is more team-oriented financial fraud.Therefore,on the basis of the past,this article proposes a more efficient and accurate anti-fraud community discovery algorithm,and builds a more advanced anti-money laundering detection system.The main research contents of this paper are as follows:First of all,to solve the problem that the traditional community discovery algorithm is difficult to apply to complex networks with high-dimensional feature information,we have improved it and introduced the concept of information entropy.The use of information entropy is a characteristic of data uncertainty measurement,which can be calculated the degree of disorder and dispersion of features in the network.When the information entropy of feature is smaller,and its confusion gets better and contains more information.And the more information that can be provided,that is,the greater the degree of variation of the feature,the greater the degree of influence in the comprehensive evaluation of the sample.Therefore,we combine information entropy with the Louvain community discovery algorithm,use information entropy to measure the weight of the edge connection of each node pair in the network,that is,the entropy weight method,and then divide it into communities.The results show that the quality of the divided communities is higher.In response to the increasingly difficult computational speed of traditional algorithms in the face of increasingly large and complex networks,we introduced the concept of connected graph cutting,that is,the entire complex network is connected by countless large and small networks.It is composed of subgraphs.Dividing community on these subgraphs one by one is equivalent to divide community in the entire complex network.If we can divide community on multiple connected subgraphs in parallel,the final time cost will be It is much smaller than doing community discovery serially throughout a complex network.Therefore,we have introduced parallelization technology to divide the community in parallel on the entire network,which greatly improves the efficiency of the division and saves a lot of time.