Abnormal Detection and Repair of Environmental Monitoring Data Based on Generative Adversarial Network and Migration Learning
Haochen Guo, Hengmei Shen · 2024
The purpose of this study is to solve the problems of outliers and missing values in environmental monitoring data and improve the accuracy and integrity of the data. Therefore, this paper describes in detail the anomaly detection model based on Gan (Generative Adversarial Network) and the data repair method based on migration learning. By constructing these two models, we can effectively identify abnormal values and repair missing values. In addition, this paper also designs a fusion model, which combines these two methods to further improve the effect of data processing. And it is verified by experiments. The experimental results show that the proposed method has achieved significant improvement in anomaly detection and data repair. Specifically, the GAN model shows a strong discriminant ability in anomaly detection, and can accurately identify abnormal values that are quite different from the real data distribution. The transfer learning method successfully uses the knowledge of the source domain to assist the data repair in the target domain, which improves the accuracy of the repair. The fusion model integrates the advantages of both, and further improves the overall performance of data processing.