Artificial Intelligence - page 13
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
You've started from the wrong side.....
Dear Kazam.... your answer is like an carbon copy... I do remember the same answer you wroted here somewhere....
If this is my first post it doesn't mean I am new here.... it could mean that I did have some work to do and now I did consider to post
however i did read you and I am pretty sure you did some work in this area especially on the right side... . so could you present some more details about your good side (having in view that i did start with left foot (I am a native lefty 
)...
The only way I can help you for now is to show you which way to go.
so what you say is that we are on the wrong way while you are on the good one? how can you tell I am in a wrong way as long as you really don't know what and HOW we want to do?
If you do this I'll tell you what to do next /before that you should be happy with the results that you get from this step/.
we already know what we want to do all the way although any good improved idea is more than welcome. the main problem we have is that we are short to people who really know AI and willing to join us and not necessary to guide us
coming back to everyone who is interesting in AI as long as no one has constant good result and don't write to us here from an exotic beach island... I think we still have to try harder by not thinking in a box
the hardest problem is to think what would you like to do with the AI module and how because there, outside is plenty of applications who are powerful enough to give good result and as I said in my first post here, is about having a very good approach on the training method and what to train.
having over 65% accuracy doesn't mean a good result. as you wrote somewhere here, I rather try to get signals for buy, sell, or do nothing instead to predict/forecast price value (it seams we do have one point in common and I bet there are plenty more).
I didn't give that example with Hertha Berlin.... just to fill more my message.
i give that example because I do believe the coach(the way you want to train and WHAT you want to train) is more important
Having over 10 years of experience in the field of evolutionary computation I can tell if a project has a chance to be successful or not.
What is more you have a very bad attitude. You don't want to learn. You just want to hold on to your plan.
Having very little knowledge and even less experience you'll just waste your time and achieve nothing.
Stop talking about "not thinking in a box" because you don't know what this box is.
I'm not saying this to offend you in any way. Just to make you stop planning and start doing something. Success comes with time, knowledge and experience.
1) how can you tell about a project can or cannot have any chance if you know so little about that project? I bet Hamilton (the F1 formula chaompion), drive way much better than me even if I do have more than 10 years experience over him:) )
2) my attitude is not bad (even if for you sound so). like you after I post, i had this feeling that you will gone say so.
3) like all of us, my time is important to me and I would do consider a great opportunity not to waste my time if you would try to put me some question in order for you to understand if is or is not a waste time. I did ask you some question and so far I didn't receive any answers.
4) any other approach different than what i have read or hear about it could be out of the box as long as what I have read don't work well.
also here you (I mean all the members) can see our first approach.
where we did design some ANN (the simple way that kazam talk about it)
we did put it together and make the first system (I only show this in order to understand that we are not so much out of this as some believe)
we did get some goo accuracy like over 60% on some pairs, we did interpret as you say like rather to forecast a price we should forecast directions.
if you look into test result xls file, you will see and those result is like after over 10000 tests (I repeat, this was our first simple approach as you recommended us
)
also some screen shoots with ANN module and our API (we shouldn't be dependent on one platform). and the first GUI application
what I am trying to tell you is like I did try to find in any of your posts something like willing to teach us the newbie
about books... well like you I do have GB's of it and anyone who would love AI I am pretty sure have
better just PM and discuss in details but trying just to tell me (us) that everything else is wrong except what you are doing....
I told you the best way to predict the movement of chaotic time series.
It's your choice what you'll do with it.
As to your project I'm not interested in knowing anything more about it. I know enough by what you already posted.
Kazam,
I strongly agree with your perspective of approaching the mechanical learning issue. I believe in starting with a structure and developing the mechanics from the structure. This approach puts me in a universal perspective where the same laws apply in all tf and instruments.
I am interested in RIPPER algorithm as (cmiiw) it is a decision making function and cannot be compared to GA as (again, cmiiw) it is a method of brutal sweetspot extraction of continuous/discrete variables. RIPPER may pave the way to learn which search space is best (mechanical relationship), where GA is functional only after the search space is determined.
Can you point us to a easy-to-understand RIPPER algorithm explanation and [possibly code? I understand and able to build MQL4 code (learning from concept and sample codes) for GA purpose to suit any variable, but have a problem with explanations containing too much mathematical equations. Prefer a layman's communication level, so to speak.
Thank you in advance for your enlightenment so far.
oR4z0r
RIPPER is an excellent choice if you are thinking about building an automated trading system [later you might also think about looking at IREP++ and C4.5/C5.0 /C4.5 is available for free, but it's much slower than C5.0/].
As to using RIPPER you don't have to implement it in MQL language. Just extract the rules and code them in MQL /you'll get a set of simple IF [...] THEN [...] ELSE [...] rules/. This way you can use J-REP, a Java based implementation of RIPPER /i think it's part of WEKA framework/.
I personally have an implementation of RIPPER in C, which I got from William Cohen [the author of RIPPER], but I can't share it without his permission. You might try emailing William and asking him for a copy of RIPPER implementation - he's a nice guy and should help you.
Some general info about decision trees and pseudo code of RIPPER can be found here
Have a look at Chapter 9.
FANN and MQL4
for those who are familiar with FANN you will find a very interesting article HERE I think Emsi, the one who wrote the article have done a very good job
enjoy it
oR4z0r
RIPPER is an excellent choice if you are thinking about building an automated trading system [later you might also think about looking at IREP++ and C4.5/C5.0 /C4.5 is available for free, but it's much slower than C5.0/].
As to using RIPPER you don't have to implement it in MQL language. Just extract the rules and code them in MQL /you'll get a set of simple IF [...] THEN [...] ELSE [...] rules/. This way you can use J-REP, a Java based implementation of RIPPER /i think it's part of WEKA framework/.
I personally have an implementation of RIPPER in C, which I got from William Cohen [the author of RIPPER], but I can't share it without his permission. You might try emailing William and asking him for a copy of RIPPER implementation - he's a nice guy and should help you.
Some general info about decision trees and pseudo code of RIPPER can be found here
Hi Kazam,
Do you know or do you have any experience with FURIA: Fuzzy Unordered Rule Induction Algorithm?(Experimental results show that FURIA significantly outperforms the original RIPPER, as well as other classifiers such as C4.5, in terms of classification accuracy.) Philipps-Universität Marburg - Knowledge Engineering & Bioinformatics : FURIA: Fuzzy Unordered Rule Induction Algorithm