CardiacScan ❤💛🖤

Table of Contents:

-Getting Started

-Training Method

-Testing the model

-Credits

How it works:

Link to the app!

The app might take a minute to start, since I am using a free version of Render.

The app uses a machine learning model (RandomForestClassifier) to predict the likelihood of a heart disease in the user. A dataset containing previous patient's health records is used to train the model.

Click the get started button, and fill in the form. Click the predict button, and the results will be shown.

Known bugs:

• To reset the app, the user has to click the predict button, and then click on the CardiacScan logo on the top-left of the navbar.
• In the results page, the About and FAQ buttons do not work for now.

Training the model:

The dataset was converted into a Pandas DataFrame object. All rows containing null values were removed. Then, the dataframe was split into training and testing samples using the 'train_test_split' function from the 'sklearn' library. A 'RandomForestClassifier' was then used with '150 estimators'. The trained model was then dumped into a .pkl format and saved for future use.

Model accuracy, precision and other statistics:

The model, when tested on the testing dataset, resulted in a precision of ~87%, an accuracy of ~88% and had a recall score of ~83%.

Also, during testing, I tried to find the leading feature that would indicate the presence of a heart disease, which turned out to be the maximum heart rate achieved, followed by the severity of chest pain in the patient.

All tests and graphics can be located in the /data_processing folder.

Credits:

Dataset from: https://archive.ics.uci.edu/dataset/45/heart+disease

Original beating heart gif from: https://toxicflame427.itch.io/