Anomaly Detection in Medical Wireless Sensor Networks using SVM and Linear Regression Models

摘要: This paper details the architecture and describes preliminary experimentation with proposed framework for anomaly detection in medical wireless body area networks ubiquitous patient healthcare monitoring. The integrates novel data mining machine learning algorithms modern sensor fusion techniques. Knowing are prone to failures resulting from their limitations i.e. limited energy resources computational power, using this framework, authors can distinguish between irregular variations physiological parameters of monitored faulty data, ensure reliable operations real time global monitoring smart devices. Sensor nodes used measure characteristics sensed is stored on local processing unit. Authorized users may access remotely as long they maintain connectivity application enabled device. Anomalous or measurement damaged caused by malicious external parties lead misdiagnosis even death patients. authors' uses a Support Vector Machine classify abnormal instances incoming data. If found, apply periodically rebuilt, regressive prediction model instance determine if entering critical state reporting readings. Using our experiments, results validate robustness framework. further discuss experimental analysis approach which shows that it quickly able identify anomalies compared several other algorithms, maintains higher true positive lower false negative rate.