Deep neural network model for heart disease prediction.
Features used for prediction were as follows -
- Age
- Gender
- Chest Pain Category
- Blood Pressure
- Number of Years as a Smoker
- Fasting Blood Sugar Level
- Diabetes History
- Family History of Heart Disease
- ECG
- Pulse Rate
The above features were chosen so that they could be measured via use of no high cost machinery and thus the overall process of heart prediction can be as low as possible in terms of money involved.
For better accuracy of prediction results from the model, the data was pre-processed with data cleaning and data normalization being performed before subjecting the data-set to training and testing.
The target labels depicted severity level of having a heart disease. The target labels spanned from 0 to 4, with 0 indicating least chances of having a heart disease whereas label 4, indicating highest chances of having a heart disease.
The given problem was modelled as a binary classification problem with labels 0, 1 and 2 being assigned a target label value of 0 and labels 3 and 4 were assigned a target label value of 1.