COVID-19-EDA

Going through the 4 GB (currently) Kaggle COVID-19 Data set to find some interesting information about COVID-19 and predict some things.

MIT License

Stars
0

COVID-19-EDA

Going through the 4 GB (currently) Kaggle COVID-19 Data set to find some interesting information about COVID-19 and predict some things.

Get the data set here: https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge This analysis is also on Kaggle: https://www.kaggle.com/jbofill/covid-19-eda-including-nlp-with-spacy

What do we know about Incubation?

I essentially searched each sentence of each body text of the articles for the word incubation and then with regex searched for digits followed by the string "day" or "days". After this, I stripped all non-numerical characters in order to preserve pharses such as 3-7. There were still some serious outliers in the 600 range, but I ended up trimming the graph before the outliers begun.

From the data I extrapolated, I arrived at an average of 12 days is how long COVID-19 should take to incubate. In hindsight, that is right around with what is common knowledge now (7-14 days).

I plan to still tamper with the regular expression to get more accurate numbers.

What do We Know about Transmission?

I used a NLP library called Spacy to create matching rules for all words that had the lemmatization of transmit and transmission. alt text Overall, there were six variations of those words combined that were found across all documents. The actual goal on Kaggle was to return all the papers that could lead to some information on transmission, so I created a chart since to see how these words were being used among the different papers to possibly help me narrow down more detailed matches since there wasn't much to visualize about the data anyways.

Probably will not do this though since my computer is not strong enough to have more precise matches.

I used some common terms used in scientific literature for ways to describe how viruses spread and searched for them in the dataframe of papers what contained forms of transmitting.

These words were: Contact, Surface, Aerosol, and Breathing

I don't believe that every instance of these words that my code finds will be in the context of transmission to be clear.

What do we know about COVID-19 itself?

I basically searched for common flu symptoms in all the documents to see if there were any matches. alt text Seems pretty standard for what is already known about COVID-19. Although sore throat seems odly low, but I only searched 'sore throat' and could've expanded that search a bit more.

Related Projects