Device Performance Anomaly Detection Method Based on Graph Convolutional Neural Network

In this article, we discuss the key task of equipment anomaly detection, which is a major challenge in modern industrial and technological fields. We propose a new deep learning framework, which is based on Long short-term memory neural network and graph convolutional neural network. By fully capturing the spatio-temporal characteristics of equipment performance data, we can capture the complex dependencies and interactions among various indicators, which are often difficult to deal with by traditional Feature engineering methods. In addition, to improve the accuracy of anomaly detection, we have also introduced storage modules and data-driven adaptive threshold strategies. This strategy dynamically adjusts the anomaly threshold based on the real-time state of the system, improving the performance of our method in dynamic environments. Our research provides new insights and tools for device anomaly detection, and the effectiveness of our proposed method has been proven through multiple comparative experiments.