Energy-Efficient Self-Supervised Technique to Identify Abnormal User Over 5G Network for E-Commerce

Within the realm of e-commerce networks, it is frequently observed that certain users exhibit behavior patterns that differ substantially from the normative behaviors exhibited by the majority of users. The identification of these atypical individuals and the understanding of their behavioral patterns are of significant practical significance in maintaining order on e-commerce platforms. One such method for accomplishing this objective entails examining the behavioral tendencies of atypical users through the abstraction of e-commerce networks as heterogeneous information networks. These networks are then transformed into a bipartite graph that establishes associations between users and devices. The Self-Supervised Aberrant Detection Model (SAD) has been proposed within this theoretical framework as a means to identify and detect users who exhibit aberrant behavior. The SSADM methodology utilizes a self-supervised learning process that utilizes autoencoders to encode representations of user nodes. The proposed method aims to maximize a combined objective function for backpropagation while utilizing support vector data description to detect abnormalities in the representations of user nodes. In summary, many tests have been conducted utilizing both authentic network datasets and partially synthetic network datasets to demonstrate the efficacy and superiority of the SAD technique, specifically within the domain of an energy-efficient 5G network.