AI tools in analytics software key in fighting COVID-19

The AI tools in analytics platforms are perhaps now more important than ever before.

Thanks to augmented intelligence and machine learning, the data being used to understand both the spread of COVID-19 from a healthcare perspective and its impact on the world economy is more informative than it could have been just a decade ago.

AI tools in analytics platforms are able to pull data from disparate sources and use it in algorithms that almost instantaneously turn the data into the models used to comprehend potential scenarios. They are sifting through the symptoms patients are reporting to their doctors, tracking mobile devices to see who might be staying at home and who isn't, and of course pulling the hard numbers from organizations like the Centers for Disease Control and World Health Organization.

The data scientists, researchers and healthcare providers fighting the coronavirus pandemic are using all of that to try to paint a more real picture of the effect of COVID-19 than just the number of positive tests.

Amir Orad

Amir Orad is the CEO of Sisense, a BI vendor founded in 2004 and based in New York City whose software, along with the platforms developed by many other analytics software vendors, is being used by various organizations working on the front line of the fight against COVID-19. Orad recently discussed how analytics software can be used to understand the spread of COVID-19 and its potential impact on both the health of millions and the economy.

In Part I of a two-part Q&A, Orad talks about the way AI tools in analytics platforms are helping to understand the effects of COVID-19. In Part II, he delves into specific examples of how organizations, both businesses dealing with the economic effects and those fighting the virus from a healthcare perspective, are using BI platforms.

What role can the AI tools in analytics platforms play in understanding and battling the spread of COVID-19?

Amir Orad: They can play a tremendous role, and is gradually playing a tremendous role. We are in the middle of fighting a disease that we don't fully understand. [Last week], there was a question in the New York Times about what the true mortality rate of the disease is, and so far we don't know if it's 1 percent, 1.5 percent or 0.1 percent. Why? Because we don't know who is really sick versus who is not and you only measure those you know, and so on.

The entire decision-making regarding the quarantine and [the economy] are all the result of a model that is built on assumptions using analytics to predict the future based on some numbers that were also derived from analytics -- from China and other areas -- so both the input, which is the assumptions about the infection rate and the mortality rate, and the output, which is the predictive model about what will happen, and then the simulations about the economy, are all using BI and analytics even before we talk about healthcare. The decision-making, the policy-making is 1,000 percent data-driven and scenario-driven.

The decision-making, the policy-making is 1,000 percent data-driven and scenario-driven.

What about the quality of the data, the fact that there could be hundreds of thousands of people who have COVID-19 but either don't realize it or can't get tested -- how does that affect how accurate the models can be?

Orad: The famous saying in big data is, 'Garbage in, garbage out.' Obviously if you don't know something it will be very difficult to predict the future. However, you can use big data analytics to correlate symptoms and doctor visits and unexplained diseases and then make an assumption. For example, there's some very famous work that Google did with the CDC on predicting flu outbreaks -- if you can identify the increase in the number of people reporting symptoms like dry cough and high fever, and you don't know if it's the coronavirus but you can compare it to the past, that is a pretty good proxy using big data analytics to see the growth in what seems to be the coronavirus over the past, even if it's not directly measured. That's all that analytics is about, the indirect analysis and decision-making and insight-generation, even when the data is not obviously clear.

That's one example. You can take other things. People are now using mobile traffic as a way to deduce another key input into the predictive model -- how many people are quarantined? That's a really important number, and you can gain that information from mobile traffic.

And of course those are AI and machine learning technologies that are recognizing and pulling that information, right?

Orad: Correct. What those machine learning algorithms do is look at typical behavior and then changes in behavior and see the anomaly, which can be that you're not moving or you're starting to move. For example, if I worked at Verizon, I could tell you in 10 minutes how many people chose to quarantine outside their house but not at a hotel. It takes 10 minutes to run the algorithm. But some countries take it to the next level and use network analysis typically used to detect terrorism and money laundering to catch people who are sick with the coronavirus but don't know it yet. The way to do it is identify people that have the coronavirus, and then identify who was in close proximity to them, and then start to understand the contamination using algorithms.

Can the importance of AI tools in analytics be quantified?

Orad: If this was 20 years ago the number of people dead would be dramatically higher because you would not have the big data, you would not have the ability to collaborate and share data as quickly, you would not have the ability to run simulations as quickly, you would not have every citizen data scientist helping fight the disease -- it would be limited only to the CDC and its equivalents. I am certain that tens of thousands, if not millions, of lives are saved directly as a result of the big data and analytics we have at our disposal today.

Editor's note: This interview has been edited for clarity and conciseness.