The Coronavirus outbreak has affected almost all of us in some ways. Healthcare systems are painfully overwhelmed, particularly in the US and western Europe, due to more cases and proactive testing. While not forgetting the plight of the patient themselves, the hardest hit is the healthcare providers, emergency services, nurses, doctors, and other frontline warriors risking their lives for others.
But there are also scientists, virologists, statisticians and data-analysts who have an equally tough job at hand – to create predictive models and derive insights which will help Governments to deploy appropriate, well-informed and strategic controls to slow down the spread; decisions such as the extent of lockdown, social distancing guidelines, managing supply chain of depleting medical accessories, extrapolating infections and predicting fatalities, sometimes shunning ideas like filling up churches for Easter!, all backed by data.
TBH, my inspiration for writing this blog was watching Dr. Deborah Birx and Dr. Anthony Fauci on the US Federal Govt’s COVID-19 response team, constantly advising citizens and sometimes subtly correcting the President, based on scientific models, and not on intuition or “gut feelings”.
With this backdrop, I wish to present a Machine Learning and data analysis project we built in CSE 6242: Data and Visual Analyticsat OMSCS, Georgia Tech. The project doesn’t attempt to solve issues related to COVID-19 (we did it in Spring’19, much before the outbreak), but I believe the idea may inspire other students or researchers in data science to do more projects particularly in the healthcare domain as the pandemic rises. For COVID-19 related data analysis, https://rapidapi.com/collection/coronavirus-covid-19 is a great API repository and it’s free.
Coming back to our paper, it is titled.
That’s quite a mouthful, and if you are short on time, please jump to a 3 mins walkthrough of the project poster that describes the project and outcomes in a concise way. The full paper, project documentation and detailed model, analysis, architecture, and source code are beyond the scope of this blog. I hope it can be made available, but at a later point in time.
Our project also has a pilot/demo implementation called – Just fit, You can perform a live behavior risk factor analysis in less than 5 minutes which compares data against 400 thousand adults in the United States to predict your chances of contracting 5 common chronic diseases.
Check your chances of contracting these Chronic diseases by answering few simple chatbot style questions
But why is the Behavior risk factor analysis important or even relevant to COVID-19?
- Prevention is better than cure, not just the Coronavirus, but also chronic diseases.
- COVID-19 related complications and subsequent death are not because of the virus itself, but because of organ failures attributed to pre-existing conditions such as Pulmonary Disease, Heart Disease, Kidney disease, and others due to a weakened immune system. These are the exact chronic diseases we predict with our model after comparing against a large number of people based on the demography, lifestyle and dietary habits of the individuals taking the assessment.
- Knowing your chances of contracting a chronic disease without going to the hospital, can ease out the currently overwhelmed health care systems.
- Data and insight are crucial to all socio-economic strategic decisions that Govt. and healthcare organizations can make today.
I would like to thank all my project mates for some great work, this credit was due for long.
Stay safe, follow social distancing, spend quality time with family and learn new things (virtually), we will overcome this!
In its current form, the intent of this blog and the demo is purely educational and it shouldn’t be used or interpreted otherwise, for example, as an alternative for health or medical advice, self-assessments, etc. However, we do believe in the correctness of the method and theory behind the process and analysis. Please use this information only to gain a preliminary knowledge of how to acquire, cleanse and build a data model, analyze the data, make predictions using machine learning algorithms and finally present the information on interactive data visualization widgets like D3.js