A Novel Approach for Anomaly Detection in Time-Series Data using Generative Adversarial Networks

Abstract

Data anomalies are found using anomaly detection. Generative adversarial networks (GANs) can produce synthetic data and learn complex patterns. Temporal dependencies and relevant features are needed to identify timeseries anomalies. A GAN is trained to detect anomalies in timeseries data. The model performs better with synthetic data. Traditional ML models struggle with the difficulties of anomaly identification in time-series data due to issues such as dealing with high dimensionality, capturing temporal correlations, and detecting infrequent events with skewed class distributions. This study tested machine learning techniques for anomaly identification in time-series data using the Yahoo! Webscope S5 dataset. In terms of F1 score (0.912), precision (0.S99), recall (0.925), AUC-ROC (0.976), TP rate (0.925), and FP rate (0.045), the suggested approach beat the baseline methods. The suggested method uses GANs to generate synthetic data for anomaly identification in time-series data.