Comparing Data Base Engines for Building Big Data Analytics in Obesity Detection

Obesity is a growing problem that has reached a pandemic dimension. Diagnosis of obesity is based on the body mass index, regardless other important indicators related to metabolic impairments. One of the main problems is the difficult identification of obese subjects or in risk of developing obesity. New health information systems supporting massive amounts of physiological, treatments and lifestyle data have been proposed elsewhere, however these have also introduced a significant data and decision overload problem. In this paper we present a comparative study on data base management engines for supporting big data analytics for the identification of obese subjects based on Electronic Health Records. We compared relational and non-relational approaches to address scalability and performance in a tertiary hospital. The experiments have evaluated data from five different hospital services on a data-mart containing 20,706,947 records from the University Hospital La Fe of Valencia (Spain). Experiments where based on data load and query with different configurations and restrictions. NoSQL approach yielded better results when compared to relational engines for all the proposed experiments