Automatic on-line risk analysis of suspicious transactions

Abstract

Portion of the electronically processed agendas and transactions of all kinds constantly increases and also the volume and importance of the processed agendas grows. Therefore, the need for supervision and inspection of these transactions is also increasing. One possibility is to accomplish control e.g. with the use of specialised mining instruments. The main disadvantage of this approach is the fact that the discovered suspicious transaction was completed long ago and it is impossible to change it. In such a case, we need to inspect and evaluate transactions in the real time still before their completion. But the complexity of real-time analysis of transaction rapidly grows with the increasing set of aspects of this assessment. This paper describes the conception, architecture and possibilities to utilize a system which contains mainly the on-line risk analysis and simultaneously it enables an automatic adaptation by the means of utilizing conclusions from a feedback analysis of mining instruments of the data warehouse. This system is suitable e.g. for the automatic on-line analysis of risks of credit card payments, authenticity of the submitted projects (at university), the submitted tax or customs declarations, etc. Without any detriment to generality, we shall use the last mentioned application domain for explaining the system properties.