Does vaccination help reduce the number of infection cases? Well, our web app Case-predictor helps you find that out for the next 100 days. It aims to provide visualization for our prediction and easy interaction to further explore our model. We wanted to gain more insights from our prediction model. Hence we went beyond the initial tasks of generating a list of raw numbers, we wanted to see a graph for the future trends which we are observed through significant statistics easily.

Instead of having to work with the package, we wanted people to experience the model itself. Hence we built a platform for people to check out the predictions through the model. Users can view the trend graph and download the prediction files easily.

How we built it

Our app is built exclusively by python scripts. With the aid of the powerful library streamlit, we are able to create a website app in 24 hours from 0 experience. The file utilizes the backend package infect_predict which we developed as a part of the data science challenge from DCP. It uses an ensemble learning model with LSTM, CNN and GRU's combined together to form an accurate model.

Challenges we ran into

Drilling through all the challenges of creating an ensemble learning model, stacked with multiple machine learning models particularly LSTM, CNN, and GRU's to establish the base of the web app. We started with 0 experience and pushed ourselves to actually write something in 24 hours. Starting with analyzing the data through EDA processes and then formulation Random forest algorithms to try out and get the feel of data we stumbled upon the vast knowledge of Data Science.

Accomplishments that we're proud of

For this app development, we are definitely proud of that we have actually launched an app in 24 hours that can successfully proceed with no bug being observed, Additionally create a next prototype for the V1 version to grow and scale our app with DCP services.

Next for Case predictor

We want to develop a suite of machine learning models with the help of hyperparameter tuning support provided through DCP compute services. Our next iteration of the app is well planned through Figma and we are excited about what more can we bring to the world and data science community at scale.

For the Figma design challenge - Link to the file

Built With

  • dcp
  • figma
  • python
  • streamlit
  • tensorflow
Share this project: