Uncompromised Accuracy: Fast and Reliable Multivariate Anomaly Detection for Satellite Signals

In the realm of multivariate anomaly detection, Deep Neural Networks (DNNs) have garnered attention. However, relying solely on a single DNN model may not achieve the optimal balance between accuracy and time efficiency. Non-linear variants of Kalman filter models (EKF, UKF) are known for their efficient time complexity but often compromise accuracy. On the other hand, deep learning-based models like Transformers and Recurrent Neural Networks (RNNs) excel in accuracy but introduce complexity challenges. This paper introduces the Selective Points Anomaly Detection (SPAD) method, which strategically merges accurate and time-efficient algorithms by leveraging a selection of multiple models. The optimal model fusion that maximizes the accuracy-to-time ratio is determined by assessing the estimated covariance from both sets of algorithms. The results demonstrate a superior Accuracy-to-Time Ratio (ATR) by at least 30% and 33% compared to the best existing method for SMAP and MSL datasets, respectively.