ekramasif/Predict_the_Diabetes

Training an Artificial Neural Network with this dataset and Predict the Diabetes.

Training an Artificial Neural Network with this dataset and Predicting the Diabetes.

About Dataset:

This dataset is originally from the National Institute of Diabetes and Digestive and Kidney Diseases. The objective of the dataset is to diagnostically predict whether or not a patient has diabetes, based on certain diagnostic measurements included in the dataset. Several constraints were placed on the selection of these instances from a larger database. In particular, all patients here are females at least 21 years old of Pima Indian heritage.

Training/Validation Split:

I use to 'Randomly' select 80% data for Training and 20% data for Validation/Test purposes. Those 80-20% split should come uniformly from each of the TARGET types, meaning I was choosing exactly 40-10% data from TARGET = 0 and the other 40-10% from TARGET = 1 (but randomly within each target).

Dataset: The dataset contains the following Columns (for clarity see the 'data.tsv'):

1) Preg = Number of times pregnant.
2) GLU = Plasma glucose concentration a 2 hours in an oral glucose tolerance test
3) BP = Diastolic blood pressure (mm Hg)
4) ST = Triceps skin fold thickness (mm)
5) INS = 2-Hour serum insulin (mu U/ml)
6) BMI = Body mass index (weight in kg/(height in m)^2)
7) DPF = Diabetes pedigree function
8) Age = Age in years

9) Outcome  = 1 - YES (meaning the patient might Diabetes); 0 - NO (the patient doesn't Diabetes)

parameters I use

   Activation functions: Sigmoid
   Optimizer: Adam
   No. of Layers: 6 hidden layer & 1 output layer
   Units: 800, 500, 250, 150, 75, 40, 1
   Loss: binary_crossentropy
   Metrics: accuracy