Inspiration

The aim of our project was to develop a risk calculator that would help in calculating the probability of being infected with Covid19 (Morbidity Risk). It would also assess the risk of death (Mortality Risk) of the person. Along with this, the project also shows the number of urban and even rural hospital beds available near the user. For this assessment, we took the user’s real-time personal data like age, height, weight, gender, etc, and predicted upon the basis of this data. We also provide dynamic stats of the coronavirus in India.

What it does

It was proven that the coronavirus was not deadly to most of the people who got affected. So, we decided to go through a dataset of past patients and make an ML model that could accurately identify whether a patient needed hospitalization or not depending on their health history and health score. This method ensures that only the patients who have a severe chance of mortality are assigned a bed while the other patients can stay and heal at home. Another area of concern that was identified was the availability of infrastructure like beds in hospitals near the user.

How we built it

We collected personal data from various websites for our platform. It ranges from normal data like height, weight, gender, age, to more medical-related data like the number of masks, sanitizers, any diseases. We even asked for their opinion as to how the user as well as their family follow social distancing and activities like washing hands. In totality, we took 33 inputs pertaining to a sample of 735 people. We used about 16 inputs out of 33 to predict the risk of being infected with Covid19 and all 33 to find out the user’s morbidity risk. Here, we used Multiple Linear Regression as our model to fit the data with the accuracy coming out to be 98% for Covid19 risk and about 93% for the mortality risk. Apart from this, our team used Vanilla JavaScript, jQuery, and Bootstrap for the frontend part. For the backend, we used Nodejs as well as Flask to connect our Machine Learning model to the web services.

Challenges we ran into

We all know how this pandemic of Covid19 has disrupted everyone’s life. It has caused the world to be at a standstill. We worked upon this problem as this is the problem of the hour which needs to be solved with utmost priority. The real-world problem our group identified was the massive overburdening of our healthcare systems. This was because once a patient was identified as being infected he/she was assigned a room and ordered to stay in the hospital regardless of the individual’s previous health score.

Accomplishments that I'm proud of

This solution provided by us is all-encompassing. It not only predicts infection and their morbidity risk but also shows the real-time status of the beds available near the user. At one platform, the users will get a complete assessment of risk and availability of medical facilities around them. Hence this solution is far better than the existing platforms as presently a user has to visit various platforms to get the same information.

What we learned

We used IBM Watson Assistant chatbot to help the user in navigating the website.
Integrating ML models to API's

What's next for COVID-Helper

Improvements: If we had access to more patient information our model could become more accurate. In the future, we would like to add more features to the Watson chatbot that we incorporated into our website.
Sustainability concerns: We feel that such a product has a good future as it helps in reducing the overload on communication systems and helps in rallying the people in times of crisis. It reduced confusion and provides an assurance and acts as an online doctor.

Share this project:

Updates