Coronavirus Prediction Flask Web Application

As a relevant topic in early 2020, it was decided to create a web application on COVID-19 pandemic as a group midterm project on class of "Web Application Development". In order to achieve a maximum grade it was also required to use a Machine Learning model in the project.

Description and Screenshots

Flask Web Application University Midterm Project, using Flask, Bootstrap, Flask-SQLAlchemy, Pytorch, JavaScript, AmCharts.

Final application allows you to:

• View landing page describing the use of application
• Sign Up, Sign In Authorize via Google
• Monitor Live Data on COVID-19 via source
• Cross-reference data over the world in one table
• Watch historic data and plots
• View information on specific countries
• Parse and display latest news with source links on COVID
• Download a simple PDF report of COVID data

It is also possible to view predictions for COVID cases and deaths on additional page.

You can watch or download project PDF presentation here: watch presentation

Deployment

Currently deployed on Heroku: Coronavirus Flask App (early version) Final application is not deployed due to Heroku restrictions on upload size of data.

Neural Network

To make predictions, Linear model was trained on PyTorch using day-to-day data. Model makes precitions for the next 10 days by default.

• In order to make more complicated, precise and sophisticated models, one can use ARIMA models or Deep Learing models like RNN, GRU, etc.

class Net(torch.nn.Module):
    def __init__(self, n_feature, n_hidden, n_output):
        super(Net, self).__init__()
        self.hidden = torch.nn.Linear(n_feature, n_hidden).to(device)
        self.hidden1 = torch.nn.Linear(n_hidden, 800).to(device)
        self.hidden2 = torch.nn.Linear(800, 500).to(device)
        self.hidden3 = torch.nn.Linear(500, n_hidden).to(device)  # hidden layer
        self.predict = torch.nn.Linear(n_hidden, n_output).to(device)   # output layer

Before running or deploying web-application, it is neccessary first to train PyTorch Linear model.

Route to train and update weighs on Cases model: /api/predict/cases

Route to train and update weighs on Deaths model: /api/predict/deaths

Models located in folder: app/models

Instructions to Setup and Run:

• Clone this repository.
• Use git clone https://github.com/yaiestura/coronavirus_prediction.git to clone this repository to your computer
• Install pip3 on your system by sudo apt-get install python3-pip if not already installed.
• Create a virtual environment by the name of venv virtualenv venv. Information in setting up virtualenv can be found here.
• Activate your virtualenv by source venv/bin/activate script
• Execute a pip3 install -r requirements.txt command to install the required packages.

Working:

• Open a terminal and enter python3 run.py
• Finally, go to localhost:5000 to display the start page of application.
• Or you can just run a bash script sudo bash deploy.sh

Future:

Final product is not ideal, and much more was expected to be done or could possibly be done:

• Application lacks smoothness, stability, lacks of tests(e.g unit tests)
• Source of the data: worldometers.info is parsed, so application may crash if HTML structure of original website is modified
• It was expected also to link COVID with Economics data, Market shifts, and to Create another Pytorch model