Grid Financial Data Enhancement and Anomaly Detection Based on Generative Adversarial Network Driving

With the increasing business complexity and digitisation of financial systems in power grid enterprises, financial data presents high dimensionality, imbalance, noise and other characteristics, which brings serious challenges to data analysis and anomaly detection. Traditional rule-based or statistical methods are difficult to effectively identify potential financial anomalies, especially in small samples and rare anomaly scenarios with low recognition rates. To solve this problem, this paper proposes a generative adversarial network (GAN)-driven approach for grid financial data enhancement and anomaly detection. The method first constructs an improved GAN model to generate high-quality synthetic financial data to alleviate the problems of uneven data distribution and insufficient samples; subsequently, an integrated anomaly detection framework is constructed by combining the information from the generator and the discriminator to improve the recognition ability of anomaly samples. The study shows that the GAN-based financial data enhancement method can not only effectively improve the model performance, but also has good scalability and practical value.