Machine learning based models for cardiovascular risk prediction

Abstract

Discovering the initial signs of the illness with geospatial information will facilitate and encourage the fact-based policies from health prospective and appropriate management approaches to be developed for diseases. During the paper, we tend to propose a machine learning primarily based prognostic modelling framework, which may run in static/low speed, massive information from electronic health records, furthermore as extreme velocity, streaming massive information settings captured from wearables, like fitness bands and biosensor watches. During our paper, we describe a scalable algorithm called Neuron network, which is used to achieve highly accurate results in fuzzy data. We have presented the outcomes of the proposed framework implementation for static and low-velocity/volume settings from the EHR & clinical DBs, with the experimental authentication of the planned framework, for 2 openly accessible CVD data sets which are 'NHANES' dataset, and the 'Framingham Heart Study' dataset), shown promising outcomes, in terms of performance of different modelling algorithms for the disease status prediction.