A regression based adaptive incremental algorithm for health abnormality prediction

Abstract

Existing learning systems for health prediction require batch-wise data or sub-text along with data to begin the learning process. These techniques are slow in learning and require more time to achieve a commendable accuracy. The techniques also provide less scope for adaptation to varying data. Since health parameters change dynamically, there is a need to reduce false positives. In this paper, a Regression Based Adaptive Incremental Learning Algorithm (RBAIL) is proposed. The novel RBAIL algorithm performs regression on the vital parameters such as Heart Rate, Blood Pressure and Saturated Oxygen Level to predict the abnormality. It also validates the data before learning, thus reducing the probability of a false positive. The proposed algorithm has been validated with varied data and is observed to provide increased accuracy in prediction and adaptability to fluctuating data. Simulation over real world data sets is used to validate the effectiveness of this algorithm.