Fraud Call Identification Based on Broad Learning System and Convolutional Neural Networks

Abstract

In recent years, fraudulent methods are constantly updated, criminal information is more hidden, and there is a problem of subjectivity of artificial feature design in traditional model feature engineering. To address this problem, a model based on broad learning and dual-channel convolutional neural network is proposed (BLS-DCCNN). First, the broad learning system is transformed from a supervised prediction method to an integrated feature generation method to generate mapped features and enhanced features for the original data. Then, the generated features are reconstructed to integrate the module to reconstruct the data distribution. Finally, a dual-channel convolutional neural network is combined with the shallow and deep layer network structure to extract global and local features, predict the final category labels, and introduce the Focal Loss function is introduced to solve the problem of positive and negative sample imbalance. Experiments and model comparisons are conducted on real telecommunication datasets, and the exper-imental results show that the model has significantly improved both accuracy, recall and F1 scores compared with traditional machine learning models such as support vector machines and random forests, and deep learning models such as long and short term memory networks.