Determination of various Deep Learning Parameters to Predict Heart Disease for Diabetes Patients
S. Narmadha, S Gokulan, Modachakanahally K. Pavithra, Rajendrane Rajmohan, T. Ananthkumar · 2020 International Conference on System, Computation, Automation and Networking (ICSCAN) · 2020
Heart disease is one of the world's major diseases which may come due to high blood glucose, high blood pressure and weight, etc. Diabetes which occurs through insulin and blood glucose level can also be the reason for getting heart disease. Deep learning model is used to prognosis the heart disease for the diabetes patients. There are two model Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). By comparing the parameters of both methods, the GRU gives the best results. Diabetes patients have their records about their diabetes, which is took as the input for predicting the heart disease for that patients. The input involves patient's id and their timing of checking. Code and Value tell about the types of diabetes. Glycemic Index value shows about the dejection of the food. Heredity shows whether sugar comes from their generation. The input also has the details about medication, blood pressure and cholesterol. Gated Recurrent Unit is the best method; it does not have any separate memory for storing the date from previous input. It uses the relevant gate to remove the unwanted data and the needed from the input. It combines the data and needed and starts predicting the results. GRU predicts with the learning rate. Learning rate uses back propagation to test and obtain the most suitable parameters for activation, loss and optimization function. The treatment of heart disease is rendered for patients by these parameters and processes.