An efficient approach for outlier detection in big sensor data of health care

Abstract

Summary In recent years, wireless sensor networks are pervasive and are generating tons of data every second. Performing outlier detection to detect faulty sensors from such a large amount of data becomes a challenging task. Most of the existing techniques for outlier detection in wireless sensor networks concentrate only on contents of the data source without considering correlation among different data attributes. Moreover, these methods are not scalable to big data. To address these 2 limitations, this paper proposes an outlier detection approach based on correlation and dynamic SMO (sequential minimal optimization) regression that is scalable to big data. Initially, correlation is used to find out strongly correlated attributes and then the point anomalous nodes are detected using dynamic SMO regression. For fast processing of big data, Hadoop MapReduce framework is used. The experimental analysis demonstrates that the proposed approach efficiently detects the point and contextual anomalies and reduces the number of false alarms. For experiments, real data of sensors used in body sensor networks are taken from Physionet database.

AUDIT: AnomaloUs data Detection and Isolation approach for mobile healThcare systems

“…The performance of Mahalanobis distance‐based approach (Salem, Liu, et al, ) is analysed over the same data set and results show that it achieves Recall of 93.80 % and incurs much higher false positive rate of 8.7 % and much lower accuracy of 92.4 % . Therefore, AUDIT and the approach of Saneja and Rani () achieve higher Recall and accuracy with lower FPR . Thus, it is suitable to combine these metrics into a single one to integrate the trade‐off between Recall and FPR .…”

“…The approach of Salem, Liu, et al () requires O ( np 3 ) time to recompute the covariance matrix with each new observation, where n is the sliding window size and p the number of physiological attributes. The approach of Saneja and Rani () can be computed in O ( nm 2 ) where m is the number of sensors. The benefits gained by our approach in term of time complexity make it useful and efficient for real‐time mobile health applications with constrained device.…”

“…We have chosen the Mahalanobis distance‐based approach (Salem, Liu, et al, ) and the approach proposed by Saneja and Rani () as baseline methods. We tried to avoid developing them from scratch and used the R Software to perform the analysis on the competing approaches.…”

“…It detects 98.4 % of all true anomalies with less than 2 % of missing detections. The performance of Saneja and Rani () is analysed over the same medical data set, and the results depict that it achieves Recall up to 96.79 % and incurs lower FPR of 0.2 % . The performance of Mahalanobis distance‐based approach (Salem, Liu, et al, ) is analysed over the same data set and results show that it achieves Recall of 93.80 % and incurs much higher false positive rate of 8.7 % and much lower accuracy of 92.4 % .…”

Standalone noise and anomaly detection in wireless sensor networks: A novel time‐series and adaptive Bayesian‐network‐based approach

“…In addition, it just tackles outliers in low‐dimension datasets, and therefore, it is not able to detect anomalies in the higher dimensions. In health care, an application of outlier detection is proposed by Saneja and Rani . This approach is based on correlation and dynamic sequential minimal‐optimization regression (SMO).…”