Covid-19 Infection Predictor

Detection of COVID infection probability using Machine Learning to Prioritize tests

Abstract The idea is to create a machine learning algorithm which will be able to detect COVID - 19 infections by looking into the data of people who have already taken the test. We can come up with a method to collect symptoms + whether the person is infected or not and train a machine learning model on top of that. This way we can prioritize the tests by pre-detecting the probability of infection using a machine learning model which has trained using the data of people who have already taken the test. For instance, 40, 000 people come for the test at a time while the capacity is just 8000. We can run these people's symptoms through a machine learning algorithm which will tell the probability of their infection and in turn we can select 8000 people with the highest probability of infection. This way the most likely infected people can be checked for the infection first and hence the spread of this deadly virus can be stopped to a large extent. I have explained and trained the model on my YouTube channel whose link has been attached with this submission.

Idea Highlights

The idea is to stop the transmission by prioritizing tests and hence detecting the cases quickly Data can be collected on the symptoms of COVID-19 A machine learning model is then trained on the data to find out the probability of a person having the infection The model is then used to find out whom to test for the infection first under a limited testing capacity The same model can be used to find potential candidates for conducting random tests under limited capacity

Machine learning model parameters

A team of doctors can sit down to find out the best model parameters. A sample set of such parameters is as follows:

Features: Average Fever - Continuous Body Pain – 0/1 – Binary Age – Discrete Runny Nose - (0/1) - Binary Difficulty breathing – Categorical : -1/0/1

Label: Probability of Covid-19 Infection

About the YouTube video A YouTube video of the solution has been attached with the submission form and is available to watch here - https://www.youtube.com/watch?v=02eZFXALcl4 Data has been randomly generated for the YouTube prototype Further a UI with a form, capable of inferring the input data from the trained model has been created and shown