The Art of Prediction

[kmguru]

Prediction is an important subject of mine, since I have to make a living out of it. The reason it is in the pseudoscience department is to explore both scientific and not so scientific methods to arrive at the results. In business, we use the prediction programs extensively for supply chain management and a whole lot of other areas.

The problem is, they really dont work in large and complex systems. It is no more accurate than the daily horoscope one reads in the newspaper. But people spend miilions in setting up computer systems. So, when Madam X tries to predict your company future, why we laugh at the prospect while forking over millions for systems that provide similar results?

In relations to Aztec's prediction and demise, are we doing any better with Enron, Terrorism, Polariod, Argentina, K-Mart and so on? Is Stochastic method any better than probability model? Is there a better one than Ouija Board or Tarot reading?

Any thoughts? Ideas?

 

[wet1]

they really dont work in large and complex systems

Let me play the dummy. Where does the present system fall apart? I am not totally familiar with this end...

 

[kmguru]

Here is an example: (severely simplified)

Simple system: Sales of #2 pencils
What would be the number of pencils we would sell tomorrow?
Answer: Use formula 5 from the folowing numbers
Last year this day plus 1, we sold 5000 pencils
The variance from last year yesterday is +100 pencils
Probability at variance is 90%
Average increase in sales is 100 pencils
So, we could sell 90 more pencils from last year = 5090 pencils

Computer predicts 5090 pencils

Complex system: Sales of inkjet printers
What would be the number of inkjet printers we would sell on 5/25/02?
Answer: Use Formula 9 from following numbers
Sales on 05/25/01 = 25
Sales variance avg. from 5/19/02 to 5/24/02 = +2
National Competition fudge factor = -1
National Economy fudge factor = -2
New model fudge factor = +2
Advertisement fudge factor = +1
And on and on
Final computed prediction = 27

Real result could be anything from 0 to 50 with probability assigned to each.

There is a lot of missing data here, some of them are:

Local market conditions
Local demand vs national demand
Change in technology
Change in use of the product
Product saturation
Local weather
Football game
Highway accident

While probability is used to incorporate some of the variables, if there are too many variables and each variable comes true at different dates, that consistently skews the results from a high optimistic calculation (no one wants to predict bad results) to a disaster reality. The variables are padded from human input to be too optimistic (like stock analysts saying it is a great buy when they email their friends that it is a dog). Put all that into a calculation - you basically get garbage. The inventory gets bloated and you have to go on sale....

 

[Mr. G]

I so much enjoy the Universe being habitually unpredictable.

It makes possible novelty, serendipity and any number of other aspects of Life being largely worth waking up for tomorrow.

 

[Stryder]

The only thoughts I have on prediction pretty much follow the same pattern of thought I've had for years. Although I have been stuck in a loop of probability for years trying to work out wether it would or wouldn't cause problems.

The subject of course is one that even Einstein mentioned, it can even be classed as the grandfather paradox, and most notibly Douglas Adams had a giant super-computer use the technique to give the answer to the meaning of life.

The understanding is, information travelling faster than the speed of light. Well that's was percievably stated.

I've mentioned many times, where one ampitude of frequency can convey information, another ampitude con convey that information faster. Now I'm sure anyone that thinks about semi or super conductors will understand that.

The main problem with such system was the understanding of what would occur when your given the change. For instance:

I'm told by the machine that I would sell 57 pencils, but when I first stocked up on pencils, I only stocked 48.
I'm told this information and I decide to get some extra pencils instock.

The problem is now, that the timeline where I was short on pencils, is no longer the timeline I'm on, the same is said about the next frame of time after that. As the messages change in accordance to entrapy.

Now this is where you could look at a bundle of entrapy parallel communications as a twirling kaleidoscopic noise, where you could follow instances to the point that entrapy fades it back into the background noise, because the more current things are stronger in force.

I suppose you could probably stare at an untuned television set and realise what I'm trying to explain (although you could question, "When the first television was turned on, did it have the untuned static we have now?")

There are 100's of things that could happen in conjunction with this sort of equipment, in the future I'll have a link that can explain this far better than now.