Bayesian regression - Has anyone made an EA using this algorithm? - page 27
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Well, if there are too many predictors with a score of less than 2, then you can't isolate useful predictors by any other means than mine. How to interpret this - I don't know. Let's not forget that predictors have an impact not only on the target variable, but also among themselves. Very often it is not only the list of predictors to be removed that is important, but also the order in which they are removed.
Clearly, the predictors are usually interdependent.
You are, I feel, leading to the use of stochastic methods like simulated annealing. I use that too. I also have a program that can quite cleverly calculate the relevance of a set of predictors (1 or more) to a target variable, and I've put both greedy and stochastic selection methods in there. Somewhere greedy works better, somewhere stochastic.
BUT! If I use decision forest (or GBM) and output the significance of variables from the trained model, I see that some of them are just rarely or never used.
Are you saying that reducing dimensionality can improve model quality (reduce fit)? Is this applicable to the decision forest?
Not surprisingly, you are not a botanist in terms of quantiles, crossvalidations and bootstraps.
Clearly, the predictors are usually interdependent.
You are, I feel, leading to the use of stochastic methods like simulated annealing. I use that too. I also have a program that can quite cleverly calculate the relevance of a set of predictors (1 or more) to a target variable, and I've put both greedy and stochastic selection methods in there. Somewhere greedy works better, somewhere stochastic.
BUT! If I use decision forest (or GBM) and output the significance of variables from the trained model, I see that some of them are just rarely or never used.
Are you saying that reducing dimensionality can improve model quality (reduce fit)? Does this apply to a forest of solutions?
SVM, ada, various trees.
Dimensionality reduction is not the goal.
Predictors relevant to the target variable are taken and a predictor selection algorithm is run on this set. What it will produce in the next window is not known: maybe it will leave all the predictors, maybe part of....
PS
ABOUT GBM. For some reason I didn't get any better results than ada...
PSPC
According to my results the most efficient predictor selection algorithms in Caret (rfe, saf, gaf). Unfortunately not for all models. Do you have any experience using them?
Let me ask your opinion, is the forex market in particular currencies really a bid and ask price (people buy and sell) or is it an automated system to take money away? Like (casino) I've seen on roulette the ball fall where it shouldn't, a magnet worked 100% sure if you want 1000% sure. I will explain why such a question, the price manages to walk in such a way that the systems I used, including myself, lost money, judging by the forum I am not the only one such, ie takes out all.
Read here: http://www.foxbusiness.com/features/2014/11/12/six-big-banks-fined-43b-in-fx-rate-rigging-scam.html
This one in particular: "The OCC also found that traders discussed actions that would potentially hurt their clients but benefit themselves and their banks and agreed not to trade in particular currencies."
Let me ask your opinion, is the forex market in particular currencies really the price of supply and demand (people buying and selling) or is it an automated system to take money away? Like (casino) I've seen on roulette the ball fall where it shouldn't, a magnet worked 100% sure if you want 1000% sure. I will explain why such a question, the price manages to walk in such a way that the systems I used, including myself, lost money, judging by the forum I am not the only one such, ie takes out all.
For starters, do not use martingale in any form. :)
And price movements (medium and long term trends) are not determined by six even very powerful banks, but by fundamental reasons - trade balance, discount rate, price index, etc.
SVM, ada, various trees.
Dimensionality reduction is not a goal.
Predictors relevant to the target variable are taken and predictor selection algorithm is run on this set. What it will produce in the next window is not known: maybe it will leave all the predictors, maybe part of....
PS
ABOUT GBM. Somehow didn't get any better results than ada...
PSPC
According to my results the most efficient predictor selection algorithms in Caret (rfe, saf, gaf). Unfortunately not for all models. Do you have experience using them?
SVM, ada, various trees.
Dimensionality reduction is not a goal.
Predictors relevant to the target variable are taken and predictor selection algorithm is run on this set. What it will produce in the next window is not known: maybe it will leave all the predictors, maybe part of....
PS
ABOUT GBM. For some reason I didn't get any better results than ada...
PSPC
According to my results the most efficient predictor selection algorithms in Caret (rfe, saf, gaf). Unfortunately not for all models. Do you have experience using them?
I use either metrics of importance in decision forest, or informational metrics (based on mutual information). In general, I believe that in decision forest type models it is not necessary to pre-select......
And be careful with Fomenko and others... They might lead you into an accidental forest)))
Yes.
And be careful with Fomenko and others... they might lead you into the wrong woods)))