Prediction of Influenza from Big Data, AI

​

◇ Prediction of Influenza from Big Data, AI

​

In January 2013, the flu began to spread in the U.S. and the death toll exceeded 100. The Centers for Disease Control and Prevention (CDC) examined deaths in 122 cities and found that 7.3 percent of all deaths were from colds or pneumonia. In response, the Centers for Disease Statistics announced that the flu epidemic has reached its peak across the United States.

   

But the funny thing is that the CDC predicted the spread of the flu two weeks earlier than the flu announcement made by the CDC. It is Google’s "flu trend." 

​

Google first analyzed the flu trend in February 2009 and contributed to Nature magazine. Google used a method to detect the level of flu trends by identifying the frequency of keywords searched by Google users around the world at that time.

   

Keywords were found in keywords closely related to the actual flu, mainly due to "flu symptoms" or "flu therapy." Big data search can now be used as a predictive and proactive instrument for how flu is prevalent in many parts of the world.

   

Recently, it has been increasing accuracy by utilizing AI. Canadian StarSup company Blue Dot currently collects data on news, livestock and animal data, mosquito-like pest conditions, international air movement data and real-time climate change from 65 countries around the world.

​

Statistics say it has two faces. It is reliable and manipulative and unreliable. There's a lot of famous talk about statistics, but if you look at just a few of them,

​

 "There are three kinds of lies. Lies, lies in the teeth, and statistics."

​

"Number doesn't lie. But liars can figure out how to use numbers.

​

"Statistics can prove anything. Therefore, statistics can prove nothing."

​

Let me introduce you to a little humor.

 

 ‘Someone asked a mathematician how much 2+2 is. The mathematician gave a curt answer of 4. The answer was so simple that I asked the statistician next to me again. The statistician said the answer is 4 at 100 percent confidence level and the margin of error is zero. 

​

The answer was so complicated that I asked the pollster next to me again how much 2+2 was. The pollster looked serious, looked around carefully, closed the window and drew down the curtain, and whispered back in a tense voice, pulling the questioner's ear.

​

 "What do you want to be two and two?"

​

​

​

◇ 晩時之歎

​

It is lamentable to miss an opportunity due to the late time.

​

​