Abstract

CASTRILLON, Omar D; SARACHE, William  and  CASTANO, Eduardo. Bayesian System for Diabetes Prediction. Inf. tecnol. [online]. 2017, vol.28, n.6, pp.161-168. ISSN 0718-0764.  http://dx.doi.org/10.4067/S0718-07642017000600017.

This study presents a Bayesian system for the early identification of diabetes Mellitus based on the analysis of some variables such as number of pregnancies, diastolic blood pressure, triceps skin thickness, body mass index, heredity and age. The proposed methodology establishes and trains a Bayesian classification system, based on samples of diabetic and non-diabetic patients. The system was tested with different patients maintaining the same ratio between diabetics and non-diabetic individuals. Finally, to detect this disease, the number of hits and errors obtained by the Bayesian classifier was compared with a specialized test. The results indicate that, considering the aforementioned variables the disease was correctly detected by the system in 87.69% of the analyzed cases. However, when the variable serum insulin was included, this percentage increased up to 98.46%.

Keywords : bayesian classifier; diabetes; training system; automated detection.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )