ARIMA-DCGAN Synergy: A Novel Adversarial Approach to Outlier Detection in Time Series Data

Outlier detection in time series data is crucial for various applications including fraud detection, system health monitoring, and predictive maintenance. However, existing methods often struggle to capture complex temporal dependencies in the presence of noise and high dimensionality. A novel adversarial approach called ARIMA-DCGAN Synergy has been investigated to handle these challenges, which combines the strengths of Autoregressive Integrated Moving Average (ARIMA) models and Deep Convolutional Generative Adversarial Networks (DC-GANs). The ARIMA component captures linear and short-term dependencies, while the DCGAN learns nonlinear and long-term patterns in the data. The proposed approach enables generating of realistic time series samples using DCGAN and the outliers detection based on the reconstruction error using ARIMA. Experimental outputs on both synthetic and original datasets exhibit excellent results of ARIMA-DCGAN Synergy compared to sophisticated outlier detection techniques. The proposed model delivers superior results and surpasses all other cutting-edge models with an accuracy of 98.81%, a precision of 98.92%, a recall of 98.97%, and an F1-score of 98.94%.