Machine learning is used to learn the properties or characteristics of any system from history as well as continuous observation of the various parameters of the system. Traditional static learning algorithms model the system during the training period with the available data. These algorithms do not provide a scope for adaptability to gradual changes in the parameters characterizing the system that is to be learned. The existing incremental learning systems are computation intensive and do not support self-correction using feedback. This paper proposes a novel incremental learning algorithm, Online Dynamic Regressive Learning (ODReL), which overcomes the disadvantages of the existing system. The proposed ODReL is applied to the prediction of the health parameters and detection of the abnormalities in the vital health parameters of a person. © Research India Publications.