Telecom fraud detection using social pattern

摘要

Generating fraud indications related to communications devices. A fraud metric is calculated for known nonfraudulent users. A fraud metric is calculated for known fraudulent users. An initial value is selected for the threshold. The fraud metric is calculated for a third set of users. The users of the third set are classified as being fraudulent or nonfraudulent according to the threshold. A false positive rate is determined and compared to a maximum false positive rate. When the rate exceeds the maximum, then the threshold is raised and the process is repeated until the false positive rate does not exceed the maximum false positive rate. A false negative rate is determined and compared to a maximum false negative rate. When the false negative rate exceeds the maximum false negative rate, then the threshold is lowered and the process is repeated until the false negative rate does not exceed the maximum.