Online Incremental Learning Algorithm for anomaly detection and prediction in health care
Kirthanaa Raghuraman, Monisha Senthurpandian, Monisha Shanmugasundaram, Bhargavi Bhargavi, V. Vaidehi · 2014
Anomaly Detection in health care by monitoring the vital health parameters of patients is a challenging problem in machine learning. The existing algorithms do not process the data incrementally and hence are not very effective in predicting the anomalies accurately and at the correct instance. In this paper, in order to process the health data in an online fashion a novel Online Incremental Learning Algorithm (OILA) is proposed. The OILA predicts the health parameters using a regression based approach with a feedback mechanism to reduce error. An alert is generated when an anomaly is seen in the health parameters, thus alerting the doctor to be cautious. The algorithm is compared with Kalman Filter for comparing the prediction capabilities of OILA with Kalman Filter. The proposed algorithm is validated with real time health parameter data sets for health parameters namely heart rate and blood pressure.