Illustration by Pablo Stanley
Words by Jack Riewe
The coronavirus, or COVID-19, is a terrifying pandemic. There are a lot of misconceptions about the disease and few are looking at actual data to provide transparency about the virus. Medical data and machine learning have an inventive relationship with each other and can work together to identify probability of diseases in patients based on data, predict appointment no-shows, and more.
For this use case, we will explore how to predict the recovery rate using a Coronavirus dataset with no-code machine learning.
The Value of Machine Learning in Disease Prevention
One of the top worries of COVID-19 is lack of databases and accessible tools. This is a great opportunity for low-code/no-code tools to be put to work. Medical professionals may lack the ML and coding expertise to get quick outputs and insights that are critical to human health. This is where no-code ML thrives.
For this use case specifically, someone in the healthcare industry could measure trends to predict where the virus was heading, which countries were most prone, and compare confirmed cases to recovery rate to draw conclusions about what factors were related to recovery without code.
With the proper data, emergency relief organizations could predict where to allocate resources, doctors could identify the characteristics of the people most likely to become infected and create personas, or countries could keep track of where the disease was concentrated and quarantine accordingly.
However, using this dataset, we will analyze these things on a basic level to inspire creativity on what you can do with no-code machine learning and medical data.
About the COVID-19 Dataset
For this use case we used a dataset that measures the daily level information on the number of COVID-19 cases across the globe. The dataset is updated every 12 hours, so we will only use data from January 22 to March 10, 2020.
The type of data we will use include:
Date of observation, Province/State, Country, Last Update, Confirmed, Deaths, Recovered.
The Analytics and Prediction Process
If you don’t know by now, we’re a no-code data science tool. We have taken the traditional time-consuming machine learning process and simplified it into 3 steps. Essentially, we’re loosely following the same process as this technical article—only without code.
With this dataset, we can upload into our platform, ask questions, and get an algorithmic output in English.
Comparatively, the traditional process of preparing and importing data would look like this:
Like we said before this process is extremely easier and more efficient for non-technical health professionals to make decisions about the Coronavirus.
We Will Show How to Predict Recovery Rate With Obviously AI, But . . .
Before we show the results, we'd like to say these outputs are based on the data provided by the dataset collected by John Hopkins University. Our predictions are only as good as the dataset provided. This is mainly to show what non-technical users can do with their data and use no-code for improving public health.
Predicting the Recovery Rate With Personas
Using the personas feature, you can predict the recovery on a country-by-country basis. It's easy to create a persona based on the countries used in the dataset. We made a persona based on the United States.
We created a United States persona and entered the today's confirmed cases of COVID-19 and the confirmed deaths taken from the New York Times. We set the observation date to the current day, the day this post was updated (3/12/2020).
All of the attributes you see are changeable to compare the recovery rate by country based on the dataset.
For example, we changed the Observation and Last Update Date to 3/16/2020 and entered a random number of 1,500 confirmed cases and 45 deaths.
Based on the dataset, Obviously AI predicted about 290 (plus or minus 36.8) of the confirmed patients will recover. You can also predict the number of confirmed cases the same way you predict recovery to compare results.
Exporting the Predicted Recoveries into a shareable spreadsheet.
You can also see the predicted recoveries in a spreadsheet and share the results, like so:
From here you can take a deeper look into the predicted recovery rate by location of confirmed cases. We color-coded the countries to easily tell them apart in the spreadsheet.
No-Code Tools Can Improve Public Health
Access to information like this through machine learning is vital to making informed decisions about COVID-19. However, the insights don't stop here. There are many other insights you can get from useable data. As we said before, this post is meant to show how to use no-code machine learning to help health professionals in a revolutionary way that might've not be available to them before low code/no code.