[PYTHON] I tried fitting the exponential function and logistics function to the number of COVID-19 positive patients in Tokyo

background

Someone has fitted the global COVID-19 infection trend to exponential and logistics functions and calculated the doubling period (https://github.com/aatishb/covid), so clone it. I applied it to the data on the number of patients published by the Tokyo Metropolitan Government.

The reason for Tokyo is that if it is nationwide data, the difference in circumstances depending on the region will be mixed into one distribution, so it is easier to apply it to the ideal model of infection explosion in a limited area. .. It is also the region with the highest risk in Japan.

The analysis method and precautions are written in the Notebook.

result

Notebook is here: https://github.com/msakuta/covid/blob/master/curvefit-tokyo.ipynb

** Based on Most Recent Week of Data **

	Confirmed cases on 2020-04-02 00:00:00 	 587.0
	Confirmed cases on 2020-03-26 00:00:00 	 212.0
	Ratio: 2.77
	Weekly increase: 176.9 %
	Daily increase: 15.7 % per day
	Doubling Time (represents recent growth): 4.8 days

** Based on Logistic Fit**

	R^2: 0.9753797360404987
	Doubling Time (during middle of growth):  12.02 (± nan ) days

** Based on Exponential Fit **

	R^2: 0.975379736914424
	Doubling Time (represents overall growth):  6.01 (± 0.38 ) days

Of note is the value of R ^ 2, which exceeds 0.97 for both the exponential and logistics functions, and can be seen as following a typical explosive growth trajectory.

Based on the number of cases in the most recent week, the doubling period is 4.8 days, the logistics function is 12.02 days (at the maximum gradient point), and the exponential function is 6.01.

When extrapolated

Extrapolating the exponential function is generally not a good idea, but if exponential growth continues, it will be around 20,000 in 30 days.

Of course, it is not easy to conclude that this will happen. The city has also issued a request to refrain from going out, and people's perceptions have changed significantly in the last few weeks. I think there is a good chance that it will converge quickly. However, looking at this graph, I have to think that it is quite a Noh weather to say "I'm just barely stepping on".

The scary thing about COVID-19 is that it has an unaware incubation period during which it may be spreading the virus. According to the Japanese Society of Infectious Diseases (https://www.kansensho.or.jp/ref/d77.html), this incubation period is 1 to 14 days, but it is the longest from taking measures to the effect. It means that there is a delay of about two weeks. Personally, I think it's too late to declare an emergency now.

In addition, the number of tests in Japan is considerably smaller than in foreign countries, so it is unknown how many undetected infected people are lurking.

What's even more difficult is that things change drastically every week, so I think there are some areas where ordinary people's senses aren't catching up. But even in Tokyo, a month later, like New York, the bodies overflowing from the hospital may be stored in the refrigerator.

Fortunately, the number of infected people is updated daily, so I would like to keep an eye on trends as much as possible.

reference

Although introduced by the original author of Notebook, the following video is highly recommended.

https://www.youtube.com/watch?v=Kas0tIxDvrg

https://www.youtube.com/watch?v=54XLXg4fYsc&t=1s