Research On The Application Of Link Analysis In Financial Supervision

Financial crimes always have close relationship with the exceptional fund flow in financial networks, some of the existing financial supervising techniques and measures are unable to monitor many hidden forms of fund flows. In view of this, make use of link analysis technique in data mining field, it can effectively detect exceptional fund flows in financial networks via analyzing the features of fund flows, such as quantities, trails and frequency.A single-transaction tracking measure was too hard to detect exceptional flows of funds for the enormous transactions. However, the problem will be solved after the introduction of link analysis. The detecting process with link analysis is given as follows: take the account as detecting objects, cluster the objective accounts, and then calculate the suspicious rank of the account according to the transaction frequency and the amount of money. At last, find the exceptional fund flows by analyzing the exceptional accounts. The entire process consists of the following two phases: phase of clustering and phase of computing suspicious rank and digging out the exceptional fund flows.Clustering is the process of discovering the groups in which the accounts transact with each other continually. Conventional data clustering algorithms identify groups of similar items in a dataset based on their attribute values only, little work in data clustering focuses on the relationships among datasets. In order to overcome the disadvantages above, firstly, define the frequency level quantificationally according to the frequency and money amount between two accounts, and use three clustering methods which are all based on the frequency level: link-based vector clustering method, it takes the vectors which denote the relationship among the accounts as its clustering object; graph partitioning-based clustering method, its goal is to partition the transaction networks such that connections within clusters are maximized and connections between clusters are minimized; association-based clustering method, it finds the maximal connected and frequent sets based on the theory of association rules, and get the clustering results in which the accounts have more stronger associations. In the phase of computing suspicious rank and digging out the exceptional fund flows, a calculation method is put forward according to the quantities of transactions, the amount of money and the money proportion which is supplied by the source account. The exceptional account is picked out according to the suspicious rank, and the exceptional fund flow is digged out.