# How does computer modeling of virus spread work?

Just wondering if anyone understands it. What kind of data / information do they need to give the computer to start with?

• cellphone mobility by google analitics, literaly
my old website http://home.hccnet.nl/c.born/ has changed to http://www.cborn.nl/zxfiles/ so just click it and select a file
• Crisis wrote: »
cellphone mobility by google analitics, literaly

Sorry I should have been clearer, I don't mean test and trace - I mean the modelling they used, that tell in advance how many people are likely to die from coronavirus.

• edited October 20
Oh... Fake news time again...

I do wonder why people believe that one must use a computer model to determine the number of possible deaths.

Is it because it's a computer model so therefore one can say it can't be trusted?

Fortunately there are even more accurate ways now than computer models to determine these numbers.

That being said, the most accurate (and easiest) way is to just stop all lockdowns and let people infect each other and die. And you don't even need computer models for it. If they did this in the UK or in the US, then over here we can predict our numbers more accurately.

Post edited by Timmy on
• Timmy wrote: »
I do wonder why people believe that one must use a computer model to determine the number of possible deaths.
You can only run the experiment once on a physical population, and it's tricky to contol all the variables.
My rubbish website including the redrawn Amstrad schematics and the new home of the Sinclair FAQ wiki.
• Essentially amalgamations of different types of modelling. Taking things like population density, migratory habits, the "R' rate etc. Of course no computer model is ever going to be entirely accurate and it's difficult to factor in how people's behaviour will suddenly change when faced with unknown scenarios.

• Numberphile covered one of the simplest virus spread modelling formulae here:
It gets a lot more complex in the actual models, no doubt, and you still need a way of calculating the variables such as infection rate etc.
• dmsmith wrote: »
Just wondering if anyone understands it. What kind of data / information do they need to give the computer to start with?

A few points:

The computer is only being used as a tool. Before computers were developed, the modelling of possible outcomes was done by hand and by using all the calculating machines and tools available at the time.

Second, the concept of using computers to number crunch data and produce statistics has been around for many, many years. Its now used in many areas. Some examples of so called predicting systems are those used for weather forecasting and for the outcomes of elections.

Third, the better the quality and quantity of data input to the system, this reduces the range of possible results. Note that this is likely to make it more likely to be accurate, but there are often very complex interactions so real life my take a different course.

In terms of the how, a system is put together that calculates the known variables, the known interactions and the known methods of reproduction. Basically everything that is known about the virus (that affects its spread) is put into the model. In the same way that everything that is known about the weather is put into the weather forecast systems.

In addition, the starting figure, or the current figure for the percentage of the population that have the virus is entered. As are various assumptions (e.g. how many times they think people go shopping per week).

Basically it’s a simulation run multiple times, with slightly different assumptions or slightly different initial data. After a set number of simulations, you then have ‘best case’ and ‘worse case’ predictions. Plus an average prediction.

You can then change some of the assumptions and then run it again to see what affect the change made.

Hence, you can then try to determine what measures to take in real life.

Of course, politics will get into it. Because different people have different views.

These systems are often criticised because the predicted outcome does not become true. Unfortunately the longer the time period that the model is set up to produce an outcome for, the more likely that in real life, something will change, something that was not catered for in the assumptions used for the model.

This is especially true for very dynamic systems. The human population being one such example. Some people react to the the news, others either don’t listen/watch/read it, or just ignore it.

In an area where I sometimes work, there is an emergency siren system, for use if one of the large chemical companies needs to warn people that there has been an incident. It’s hard to predict who will ignore it and who will leave the area (as recommended).

And so it is with trying to deal with a virus like the COVID19 Corona virus.

The so called survival instinct in humans varies in each individual. Everyone subconsciously as well as consciously does their own risk analysis. Unfortunately for humans (just like most mammals, birds and similar creatures), this is more effective if an immediate risk is perceived. If the timespan is too short, or too long, or it’s not possible to see, hear, smell, taste or otherwise perceive the problem, some (most) will not take any action.

Nuclear radiation is a good example of this. By the time you get symptoms of radiation damage to your body, it’s too late, a lot of damage has already been done.

This further makes models harder to develop, as the reaction of people is very complex...

Well after that, how about a game. How about game of life?

Mark
• Numberphile covered one of the simplest virus spread modelling formulae here:
It gets a lot more complex in the actual models, no doubt, and you still need a way of calculating the variables such as infection rate etc.