چكيده به لاتين
Detection of errors in the field of finance is one of the most challenging issues in studies of data mining. Due to the large volume and complexity of data accounting when we faced with this task, it is impossible to do it manually. Since many fraud detection systems have serious limitations, to identify different types of fraud in the field of after-sales services, there may be multiple systems is required with the methods, parameters and different databases with specific features.
Finding after-sales service daily fraud of vehicles between the massive amounts of network services that provide after-sales service is very difficult. To serve an after-sale service record as a fake one, a fraud should be recognized. Moreover, nature of fraud is changing in the course of time, and its type is unknown, finally these changes may involve more than one type of service. In addition each customer or agencies can cause fraud.
In this thesis a system design and implement for reducing abuse of the financial fraud Emdad Khodro Iran (IKCo) company by both customers of Emdad Khodro Iran (EKI) and/or the authorized agents of Iran Khodro. In this system, an algorithm try "to analyze the suspicious client/agency", then checkingout and paying their costs to ensure that the accuracy of the Emdad-ID is the key value for verification. In this algorithm, by useing of black lists for each client/agency, two suspicious score considering to the characteristics of each entity are taken into account. After issuance of the Emdad-ID, the detection process starts and according to customer records, suspicious score is calculated and updated. The proposed system, with one survey response data in real time, able to explore the relationship between different Emdad-IDs using the white list is frequently updated. The results of tests on a large scale agenda indicates the proposed method, is successful in detecting fraud and prevent financial abuse.
