Automatic Fraud Detection In Call Center Conversations

Abstract

In this paper, a machine learning system that automatically detects fraudulent call center conversations is presented. The system first transcribes the call center telephone conversations into text using a speech recognition engine and then it automatically detects the fraudulent conversations by a text-categorization algorithm using the transcribed texts. Several classifiers that use different document vectorizers are trained, tested and their performances are compared. The best results are obtained by using deep convolutional neural networks that use word embedding vectors as their inputs. With these networks, 43% of fraudulent calls can be automatically detected with 62% precision.