Outlier Detection Applying an Innovative User Transaction Modeling with Automatic Explanation

We present a method to detect outlier or exceptional transactions records applying an innovative user modeling. We use a large financial database to validate our method. Our method has two stages. The first stage is for user transaction modeling and it obtains user behavior according to historic transactions based on categorical or numerical attributes. The second stage is the monitoring where a new transaction is compared against the corresponding user model, in order to determine if this transaction is unusual (no standard, fraudulent or suspicious). The novelty of this method is that it provides to the user with an automatic explanation about the exception level of the new transaction (e.g. transaction normal, abnormal, suspicious, etc.). And also provides the percentage of ownership to them. According to the experiments conducted with a very large financial database, encouraging results were observed in the field of applied Business Intelligence, in particular to the financial frauds detection and in general to the outlier detection area.