An intelligent diabetes prognosis tool

Abstract

Diabetes is highly ranked among the leading causes of mortality in the Uganda [1]. There are quite many types of diabetes, which include but are not limited to Type 1, Type 2 and Gestational Diabetes. Diabetes is found in every population in the world and in all regions, including rural parts of low- and middle-income countries. WHO estimates there were 422 million adults with diabetes worldwide in 2014. The age-adjusted prevalence in adults rose from 4.7% in 1980 to 8.5% in 2014, with the greatest increase in low- and middle-income countries. This is greatly attributed to the change in people’s lifestyles like adoption of poor eating habits, sitting for long hours at places of work and increased stress levels. WHO also explores other causes of persistently elevated blood sugar levels to include infections, surgical procedures, and genetic conditions among Availability and distribution of the collected records on the different types of this disease in Uganda is very unbalanced, making it very hard to manage diabetes. Herein, we propose an ensemble- learning framework of ANN and LSTM models, to build end-to-end Intelligent Diabetes Prognostic tool to assist medical practitioners in making diagnosis decisions. The use of an ensemble framework was adopted to overcome overfitting challenges that are very common with ANN. Results showed that the ensemble model had the best classification performance with accuracy between 91% and 96% for the different types of diabetes disease. These results are competitive and very promising for Medical diagnostics.