An Incremental Identification Method for Fraud Phone Calls Based on Broad Learning System

Abstract

With the continuous development of the communication industry, more and more fraud calls appear in the user’s daily life and the crime of telecom fraud is growing rapidly, causing huge losses every year. Traditional fraud detection methods are less flexible and they all belong to passive interception and rely on intelligent terminals. At present, a more accurate and timely method is needed to deal with the evolving fraud. Therefore, this paper proposes an identification method for fraud phone calls based on Broad Learning System (BLS). We processed the text data of fraud phone calls through the first 15s of the call content identification monitoring, constructed the TF-IDF model, then converted it into a neural network based on the BLS and identified the fraud phone calls on this model. At the same time, the model can be updated quickly by corresponding incremental learning algorithm without retraining based on the BLS, which is suitable for fraud identification systems with few data features but high real-time prediction requirements. The method mentioned above is experimented and analyzed in detail. The results show that this method has higher accuracy and excellent training speed on fraud data. Compared with the original fraud identification methods, it can actively intercept and has higher accuracy. Compared with other neural network algorithms used in fraud system, the method has better training speed, can ensure the accuracy and timeliness of online fraud identification and help quickly identify fraud phone calls.