Unsupervised credit card anomaly detection method based on multi-dimensional feature tensor

• Main Authors: LIU MIN, JIN YUQING, WANG XIAO, XU HAO, FANG NING, WANG YIHAN, CHEN YANG, LIN FAN, MA XUEHUAN, DUAN MINGJIANG, CHEN GUIHUA, YAO YI, SUN WANQI, SUN LIJUN
• Format: Patent
• Language: Chinese; English
• Published: 13.09.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses an unsupervised credit card anomaly detection method based on a multi-dimensional feature tensor, which organically combines multi-dimensional multi-scale feature tensor feature construction, a multi-dimensional attention convolutional network and a recoding generative adversarial network for the first time, and generates a high-quality generation result by the multi-dimensional attention convolutional network. And the multi-dimensional and multi-scale feature tensor is coded, decoded and re-coded. The characteristics of the abnormal transaction samples are expressed to the maximum extent, and high-quality reconstruction representation is obtained; 3 sigma abnormal scoring based on time, space and category is carried out on the reconstruction features, abnormal voting is carried out based on different scales, the abnormal scales are abnormal, the noise influence is avoided, and the robustness is improved; and a streaming index calculation engine is adopted for cooperative work, so that