Efficient Anomaly Detection in Correlated Sensor Data Streams Using Living Correlation in Resource-Constrained Environments

This research proposes using Living Correlation for anomaly detection in correlated sensor data streams. Traditional correlation methods are computationalliy heavy, requiring repeated Z-Score calculations from time N = 0 to the present. In contrast, the Living Correlation adopts a recursive model, updating correlation dynamically using the previous step and current Z-scores. Experiments show that it rapidly converges to traditional correlation values while detecting anomalies 96-385 times faster, making it ideal for real-time, resource-constrained systems like industrial machinerey, oil rigs, and autonomous vehicles.