Testing real-time forecasting systems - page 28

 

It's my turn to post the first day of testing:


 

I would have given a prediction too, but only MASHKI FALLS )))))))))))

УБИТЬ ВСЕХ ЧЕЛОВЕКОВ )))

 

So the cartoon characters meet. What's "broken" about it, you only know how to bend. :о) And what does it all mean, I mean - what's the point of so many lines?

 
grasn >> :

So the cartoon characters meet. And why is it "broken", you only know how to bend. :о) And what does it all mean, I mean - what's the point of so many lines?


Yep... )) especially in my sleep bending Rodriguez bender.


Doesn't mean anything. I just don't know how serious the topic is... Just thought I'd throw in a little humor.

And in general 3 colour mashas 24, 182, 600 Very well behaved. (Green is buy red is gone)

Punctuation is the most global trend.

It's just that sometimes artifacts come out like this. I should write to the author of this TC to fix it. Found it in kodobase, it's even fresh.

 

I see you like to play with parabolas, Professor.)

Still inventing? ))

 

Somehow it came out like this:

Or like this:


 

I've made some improvements to my GRNN. The accuracy of the predictions has improved slightly. Looking into the future, we get the following prediction


 
fozi >> :

And I see you Professor like to play with parabolics ))))


No, it's not parabolic, it's pure statistics.

Still inventing? ))

This system is based on a "quantum neutrino" field (C) :o)

 
gpwr >> :

I've made some improvements to my GRNN. The accuracy of the predictions has improved slightly. Looking into the future, we get the following prediction


If it is not a trade secret, what has been finalised?

PS: I liked the non-linear AR prediction better, maybe - it's a very promising direction. But apparently identifying such a model is very difficult.

 
grasn >> :

If it's not a trade secret, what has been finalised?

PS: I liked the non-linear AR prediction better, maybe - it's a very promising direction. But apparently identifying such a model is very difficult.

The refinement consisted of removing pattern price smoothing and increasing sigma (reducing cluster size) so that the closest patterns would influence the prediction.

Reason: