A simple Diabetes Prediction System using Densely Connected Neural Networks. The Dataset is provided in the repo. The dataset has been cleaned.
Dataset Description: Glucose - Fasting blood sugar HDL - HDL or good cholesterol Chol/HDL - Ratio of total cholesterol to good cholesterol. Desirable result is < 5 Age - All adult African Americans Gender - 162 males, 228 females Height - In inches Weight - In pounds (lbs) BMI - 703 x weight (lbs)/ [height(inches]2 Systolic BP - The upper number of blood pressure Diastolic BP - The lower number of blood pressure Waist/hip - Ratio is possibly a stronger risk factor for heart disease than BMI Diabetes - Yes (60), No (330)
This data set came from Vanderbilt but the original data came from a study of rural African Americans in Virginia. I elected to delete 13 patients who were missing data. Also, the diagnosis of diabetes was based on a glycohemoglobin of greater than 7, however current guidelines use a glycohemoglobin of 6.5 or greater to diagnose diabetes.
I opted to create a Total Cholesterol/HDL Cholesterol ratio and a Waist to Hip ratio as both are now commonly used but were not part of the original data.
REFERENCE: Williams JP, Saunders JT, Hunt DE, Schorling JB. Prevalence of coronary risk factors among rural blacks: A community based study. Southern Medical Journal. 1997;90:814-820