Predict Patient’s Mortality Based On Physiological Measurements | Sample Assignment

Pushkar Nandgaonkar
Jun 21, 2022
3 min read

Aim

Aim of this project to train a model that can predict patient’s mortality based on four physiological measurements for 48 hours after admitting to ICU.

In train folder, you are given four csv files (Heart.csv, Temperature.csv, Respiration.csv, Glucose.csv) containing four physiological measurements for the same 2000 patients admitted to ICU in a hospital. Rows represents patients and the columns represent the physiological measurements of the patient taken hourly. There are 48 columns indicating the first 48 hours after the patient is admitted to ICU. 50% of patients survives the ICU and the remaining 50% die at some point after the first 48 hours.

The class labels are stored in y_train.csv (1: survives, 0: death). y_train.csv has only one column that indicates the class labels. Assume that order of patients in y_train.csv and the measurement csv files are same.

First, you need to load these four files to a numpy tensor x_train of shape (2000,48,4) which in the form of (# of training samples, # of time steps, # of features). Here # of time steps is 48 and # of features = 4 (four physiological metrics). This can be done in different ways. One way to do is to first create an empty x_train tensor of shape (2000,48,4). Load each of the csv file and simply copy data to x_train along the final axis. For example x_train[:,:,0] can store the heart measurements, x_train[:,:,0] can store the Temperature measurements and so on. Also, load class labels to y_train tensor of shape (2000,)

In test folder, you are also given four csv files (Heart.csv, Temperature.csv, Respiration.csv, Glucose.csv) containing four physiological measurements for the same test 400 patients. You will need to load these four files to x_test tensor of shape (400,48,4). The class labels are stored in y_test.csv (1: survives, 0: death). Also, load class labels to y_test tensor of shape (400,). y_test.csv has only one column that indicates the class labels. Assume that order of patients in y_test.csv and the measurement csv files are same.

Model:

Create a keras CNN model that contains the following layers

1D CNN layer
ReLU activation layer.
1D MaxPooling layer of window size = 4, stride = 1.
Flatten layer.
Output Dense Layer with one node and sigmoid activation.
Use binary cross entropy as the loss function
Use early stopping criteria, with patience = 3.

Hyperparameter optimization

In the model above, perform hyperparameter optimization for the following hyperparameters. The different options for each hyperparameter to be explored are shown. Only one method of hyperparameter optimization is necessary.

Kernel length = 3, 5.
Number of kernels = 6, 32, 64, 92.
Batch size = 20, 30, 50.
Optimizers = Adam, SGD, RMSprop.
Learning rate = 1*lr, 0.1*lr, 10*lr, where lr is the default learning rate of the optimizer in Keras.

Grading rubrics

Incorrect x_train or y_train or x_test or y_test: 30 points deduction
Incorrect set up for model architecture: 20 points deduction
Incorrect dictionary for hyperparameter optimization: 20 deduction
Incorrect training procedure: 20 points deduction
No Prediction on testing data: 10 points deduction

Get solution of this project at affordable price. please send your query at contact@codersarts.com we'll share you price quote

How can you contact us for assignment Help.

Via Email: you can directly send your complete requirement files at email id contact@codersarts.com and our email team follow up there for complete discussion like deadline , budget, programming , payment details and expert meet if needed.
Website live chat: Chat with our live chat assistance for your basis queries and doubts for more information.
Contact Form: Fill up the contact form with complete details and we'll review and get back to you via email