基于银行账户亲密度网络推理的团伙预测研究

Reasoning method for predicting crime partner by using intimacy network of bank accounts

近年来,针对涉众型非法金融活动在资金交易规律的研究引起了研究者的高度关注。为解决利用银行交易数据进行异常账户犯罪团伙主动发现的问题,提出一种基于银行账户非对称亲密度网络的团伙预测方法。首先,建立银行账户交易通用网络模型,将时序交易数据嵌入网络结构中。然后,利用节点的直接和间接交易关系信息,提出一种账户非对称亲密度计算方法。最终,利用节点在亲密度网络上的非对称交互信息,得到节点的异常倾向性指标。在包含传销团伙的真实数据上的实验结果表明,基于亲密度网络的团伙预测方法能有效发现潜在传销人员。

In recent years, research on the law of capital transactions in the case of stakeholder-type illegal financial activities has attracted the attention of researchers. In order to solve the problem of using the bank transaction data to actively discover the abnormal account criminal crime partners, a crime partner prediction method based on the asymmetric intimacy network of the bank accounts was proposed. Firstly, a general network model for bank account transactions was established to embed time-series transaction data into the network structure. Then, using the direct and indirect transaction relationship information of the node, an account asymmetric intimacy calculation method was proposed. Finally, using the asymmetric interaction information of the nodes on the intimate network, the abnormal tendency index of the nodes is obtained. The experimental results on the actual data of the multi-level marketing group show that the crime partner prediction method based on the intimacy network can effectively find potential pyramid shelling members.