Analysis of E-commerce Security Protection Technology Based on YOLO Algorithm Optimized by Lightweight Neural Network

Aiming at the security of logistics and warehouse management in the e-commerce environment, an improved model based on You Only Look Once version 7 and GhostNet is proposed for abnormal behavior detection. The study analyzes potential security risks in e-commerce warehouses and concludes that monitoring abnormal warehouse behavior can effectively ensure asset security, improve operational efficiency, and reduce loss risks. A lightweight object detection network is constructed by introducing efficient Ghost Module and depthwise separable convolution to optimize model performance and improve inference speed. The experiment used the Awake dataset for performance validation. The average precision was 86.8% and 83.2% in the training and testing sets, with recall rates of 77.6% and 76.5%, respectively, with significant advantages compared with the control model. In addition, the mean absolute error was 0.0023, and the area under the ROC was 0.85, indicating its efficiency and reliability in abnormal behavior detection. Therefore, the model provides an effective real-time monitoring solution for e-commerce warehouse management.