Can we predict the unpredictable?

[robin gras]

Time series forecasting is of fundamental importance for a variety of domains including the prediction of earthquakes, financial market prediction, and the prediction of epileptic seizures. We present an original approach that brings a novel perspective to the field of long-term time series forecasting. Nonlinear properties of a time series are evaluated and used for long-term predictions. We used financial time series, medical time series and climate time series to evaluate our method. The results we obtained show that the long-term prediction of complex nonlinear time series is no longer unrealistic. The new method has the ability to accurately predict the long-term evolutionary trend of stock market time series and successfully predict financial crises several weeks in advance, and it attained an accuracy level with 100% sensitivity and specificity for the prediction of epileptic seizures up to 17 minutes in advance based on data from 21 epileptic patients. Our new method also predicted the trend of increasing global temperature in the last 30 years with a high level of accuracy. Thus, our method for making long-term time series predictions is vastly superior to existing methods. We therefore believe that our proposed method has the potential to be applied to many other domains to generate accurate and useful long-term predictions.

 

More information here.

[elfmotat]

Cool stuff .

[robin gras]

Thank you!

[imatfaal]

Seems fascinating and most impressive - thanks for sharing and congratulations.

 

Seeing your work on EEGs have you considered looking at ECardioGrams and attempting to map the progress from a cardiac arrhythmia to lethal fibrillation?

[robin gras]

Thank you very much!

 

Yes, we thought about that but we have not tried yet. It is in our to do list ;-) Effectively, we expect that it will also works pretty well.

[Sensei]

The new method has the ability to accurately predict the long-term evolutionary trend of stock market time series and successfully predict financial crises several weeks in advance,

 

If your program would predict crisis several weeks before crisis, and everybody would use your, or similar programs for prediction, crisis would happen in the same moment at which program predicted crisis, as everybody would start selling in the same moment..

[robin gras]

Hello Sensei,

 

We are not specialist in finance. We do not know what the best way to use such kind of predictions is.

We only propose a method and demonstrate what its possibilities are.

However, the computational time of the method is very small (few seconds to few minutes). It means that even if the prediction of the method leads to some behaviors in the market that change completely the properties of the time series, new predictions taking into account the new state of the market could be done in real time giving some feedbacks about the effects of the reactions and giving the possibility to adjust the reactions and the behaviors of the financial institutions.

 

Robin

[Dekan]

I can see how your "time series" method, might be applicable to earthquakes. Or epileptic seizures. These both depend on strictly physical events.

The physical events are - in the case of earthquakes - changes in terrestrial sub-surface rock/magma distribution, flow, pressure and so on. Which eventually cause a seismic event.

 

And it's possible to conceive, that there might be analogous physical changes occurring in the brain. Which eventually cause an epileptic fit. And can be predicted.

 

This is OK so far. But, I can't see how the behaviour of the Stock Market can be predicted. The "Stock Market" isn't a physical thing, like a human brain, or the Earth's crust.

It's just a kind of abstract, collective noun. It encompasses millions of people, trading stocks and shares. Which are, in themselves, non-physical. "Stocks" and "shares" have no physical existence, except as numbers on a video screen, or numbers written on a piece of paper. They mean nothing, except what people agree to believe in them.

 

They are thus mere figments of human imagination. And I don't see how our imagination can be "predicted" by any method.

But - this may be unjust. Could your method be put to a test. Such as predicting the Dow-Jones Index a month in advance. At close on Oct 31, it stood at 17,390.

 

What will the Dow-Jones Index be at close on Nov 30? Can your method supply an accurate prediction?

Edited November 2, 2014 by Dekan

[Sensei]

They mean nothing, except what people agree to believe in them.

 

Not exactly.

When you've shares on your account, you're co-owner together with thousands or millions of other people.

Suppose so company issued 1 mln papers. You bought 10,000 shares. So you've 10,000 / 1 mln = 1% of company.

When you will exceed 5% of shares, any operations, selling or buying has to be reported to company and it'll inform the market in official statement within 2-3 business days. (In different countries there might be other regulations regarding significant operations on shares)

 

After exceeding certain percentage of shares, you're allowed to be included to Supervisory board by yourself or your man, or even became CEO.

 

I have stock markets accounts, and all the time during session watching prices on 2ndary monitor. I have also written couple applications.

 

[robin gras]

Hello Dekan,

 

For sure stoke and share exists and are not just "figments of human imagination". Moreover, information on the state of the world and the worldwide economy is well broadcast and directly influence the people behaviors. That means that stoke market do not vary completely randomly and abruptly making them potentially predictable even if it is difficult to do.

 

Robin

[Bill Angel]

How's this for a scenario?

Current lower prices for gasoline results in consumers having more money to spend on American products and services.

A greater demand for these products and services results in fuller employment and upward pressure on wages, resulting in an increase in inflation (via higher prices).

These inflationary pressures cause the Federal Reserve to start raising interest rates.

What happens when the Federal Reserve starts raising interest rates? THE STOCK MARKET SELLS OFF!

So now could be a good time to start selling off one's stock holdings, perhaps even short selling stock in some of the high flying glamour companies.

I'm uncertain as to whether a prediction of stock market behavior based on complex nonlinear time series could either confirm or refute this kind of market analysis.

[studiot]

In the UK financial products are required to carry the following 'health warning'.

 

'Warning any indication of past performance is no guarantee of future performance'

 

Any analysis such as yours can only undertake an analysis of past performance.

That is not to say you have have not pushed the level of sophistication forwards, I think you have so well done.

 

But all you can do when using the past to predict the future is assume that the same processes that produced the past results continue to act in the same way and no new unexpected events or processes occur.

 

You can hope to reliably predict the course of a heart condition, but cannot predict that the patient will will out of the clinic and be run over by a bus.

 

I think this is what Dekan was getting at in his comments. New events or processes are more likely with the stock market than medical or geophysical conditions.

 

So yes, progress is being made.

[Ophiolite]

Dekan has overlooked the fact that the millions of persons and institutions who hold stocks behave in a narrow range of predictable ways in response to events. (Uncharitably, most people are sheep.)

 

Or, from a different perspective, a fault plane is made up of billions of crystals and grains in contact with each other, within a stress field. That is ultimately more complex than the few millions of stockholders globally.

[studiot]

Dekan has overlooked the fact that the millions of persons and institutions who hold stocks behave in a narrow range of predictable ways in response to events. (Uncharitably, most people are sheep.)

 

Or, from a different perspective, a fault plane is made up of billions of crystals and grains in contact with each other, within a stress field. That is ultimately more complex than the few millions of stockholders globally.

 

The aggregated (inter)actions of vast numbers of participants over a wide range of inputs or conditions is the stuff of modern Science and Engineering and the results pretty impressive.

 

But this thread is about 'predicting the unpredictable', which implies inputs or conditions outside the usual or studied range which include one-off Titanic like events or walking out of a clinic and being hit by a bus.

 

Both of these events are conceivable beforehand, but are their occurrence predictable?

 

Anothe situation might be events which are not conceivable beforehand, such as the Comet disasters in the 1950s. Predicting these would have required an evaluation of the question 'What happens if we get our sicence wrong due to factors we do not know about?'

Edited December 5, 2014 by studiot