Price Prediction Based on the Periodicity Transform

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

A little bit of fun with numbers:

If we look at the crude oil prices since Septermber 2003 (NYMEX), we can make the following observations:
- prices are on a upward linear trend
- prices fluctuations are somewhat periodic;

A new signal processing tool has been proposed recently by William Sethares called the Periodicity Transform:

Jack · Dark Lord Joined: Aug 11, 2004 Posts: 5088

Fascinating! Thanks for doing the analysis.
_________________
Dieoff. Fun to watch. Better with hot buttered popcorn! new_popcornsmiley

geoman · Coal Joined: Jun 17, 2005 Posts: 9

So according to your analysis, we are at a peak in prices right now. It looks lke prices will go back down to the low to mid fifties in the next couple of months then go back up.

It also looks like your model underestimated the price of oil a few months back. Could you explain that?

VMA131Marine · Tar Sands Joined: Jul 05, 2004 Posts: 58

Another comment: It appears that the amplitude of the fluctuations in price is increasing as the price gets high. However, your prediction keeps the amplitude constant. I would agree that it is not reasonable to let the amplitude grow without bounds but what will be the limiting factor in th future?

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

Tomorrow I will try to inject some scaling factor on the predicted fluctuations maybe based on a fit on the local variance of the residuals.

There is also a possibilty that the trend is in fact slightly exponential but the factor in the exponential is very very small which makes its estimation a little bit unreliable.
_________________
______________________________________
http://GraphOilogy.blogspot.com

RonMN · Posted: Thu Jul 07, 2005 7:40 pm Post subject:

I don't mean to sound "flip"...

But do you really think that mathmatics can calculate human nature?

Whether we will "band together" or "fight a mad max senario" is beyond me...and i do believe it's beyond mathmatics as well.

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

EnergySpin · Fission Joined: Jun 25, 2005 Posts: 2381

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

EnergySpin · Fission Joined: Jun 25, 2005 Posts: 2381

I feel stupid .... Shocked

Checked my HD for papers I am randomly selecting ... apparently hyperexponential kernels are used to fit "long tailed data".
So .... they should be good for at least the part of the curve prior to depletion (then it is down the cliff) Shock

BTW thanks for alerting me to this SP methodology. Has it been tried to denoise Black-Scholes stochastic processes?
Sounds pretty close to a wavelet based SP approach.

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

MicroHydro · Light Sweet Crude Joined: Apr 10, 2005 Posts: 1286

The baseline data represents a period of economic expansion with rapidly increasing demand and slowly increasing supply. Neither might be true in the near future. I would guess that both a recession (falling demand) and oil depletion (falling supply) are near at hand. Not clear which will come first, so impossible to predict future market conditions.
_________________
"The world is changed... I feel it in the water... I feel it in the earth... I smell it in the air... Much that once was, is lost..." - Galadriel

khebab · Moderator Joined: Sep 27, 2004 Posts: 935 Location: Canada

BabyPeanut · Posted: Thu Jul 07, 2005 9:40 pm Post subject:

Based on your chart the price will come down in August. Fat Chance! Laughing