forex strategy

alex_alen

Hi dears Would anyone possibly show me an idea how to use rapid miner for testing for example some strategy trading in forex?
I really have no idea how to do that. Could you please elaborate your explanations with some example? at the moment I have no skills of programming , in such a way that I want just work with Statistics and rapid miner and some strategy that it could be tested with Statistics. 

Thanks

hughesfleming68

Thanks @mschmitz

@alex_alen. I think you need to be clear about what problem you are trying to solve first. Rapidminer is not a back testing tool. You will need that as well and a strategy to go with it. Machine learning can work well to supplement parts of an existing strategy. You might have to adjust your expectations.

I would still do the time series tutorials and Thomas Ott of neuralmarkettrends.com still has some the best introductory tutorials for time series forecasting that are well suited to beginners. They are for an older version of Rapidminer but the concepts are the same. I would also look at the Rapidminer academy which is very up to date.

Since you are starting from zero, I suggest taking a look at a book by Stefan Jansen. "Machine Learning for Algorithmic Trading". It is very comprehensive but it is not light reading. You will also need to have some experience managing risk. Things happen in the real world that no amount of testing or ML will prepare you for.

Stefan Jansen's book should scare you and for a good reason. There are no simple answers to your questions without considerable commitment and time.

regards,

Alex

hughesfleming68

One more thing....take a course in probability.This whole business is about flipping a weighted coin over and over and managing risk. Market regimes change constantly. It is likely that with ML you will be able to model a particular regime and it will work for a while....just not very long. Case in point...about 8-10 years ago, I built a model and then re-did it in Rapidminer 5.3 and wrote some tutorials for forex traders. It worked surprisingly well. I looked at it again a year ago and its predictions were only slightly better than random. Markets had changed.What also has changed in 10 years is the number of people in institutions and hedge funds that are using ML for signal discovery. All the simple edges are gone so you have to dig deep now to make something that works consistently. The game has changed considerably.

hughesfleming68

If you really must try, build a process that predicts the next day high and low. Open and closing prices are practically random but high's and lows are usually influenced by recent historical prices. Calculate the difference which will be the next days range. See if your results are better than using simple moving averages for the same purpose. Learn how to properly label financial time series. There are chapters in books written just about that.

MartinLiebig

Sounds like something for @hughesfleming68

hughesfleming68

Thanks @mschmitz

@alex_alen. I think you need to be clear about what problem you are trying to solve first. Rapidminer is not a back testing tool. You will need that as well and a strategy to go with it. Machine learning can work well to supplement parts of an existing strategy. You might have to adjust your expectations.

I would still do the time series tutorials and Thomas Ott of neuralmarkettrends.com still has some the best introductory tutorials for time series forecasting that are well suited to beginners. They are for an older version of Rapidminer but the concepts are the same. I would also look at the Rapidminer academy which is very up to date.

Since you are starting from zero, I suggest taking a look at a book by Stefan Jansen. "Machine Learning for Algorithmic Trading". It is very comprehensive but it is not light reading. You will also need to have some experience managing risk. Things happen in the real world that no amount of testing or ML will prepare you for.

Stefan Jansen's book should scare you and for a good reason. There are no simple answers to your questions without considerable commitment and time.

regards,

Alex

hughesfleming68

One more thing....take a course in probability.This whole business is about flipping a weighted coin over and over and managing risk. Market regimes change constantly. It is likely that with ML you will be able to model a particular regime and it will work for a while....just not very long. Case in point...about 8-10 years ago, I built a model and then re-did it in Rapidminer 5.3 and wrote some tutorials for forex traders. It worked surprisingly well. I looked at it again a year ago and its predictions were only slightly better than random. Markets had changed.What also has changed in 10 years is the number of people in institutions and hedge funds that are using ML for signal discovery. All the simple edges are gone so you have to dig deep now to make something that works consistently. The game has changed considerably.

hughesfleming68

If you really must try, build a process that predicts the next day high and low. Open and closing prices are practically random but high's and lows are usually influenced by recent historical prices. Calculate the difference which will be the next days range. See if your results are better than using simple moving averages for the same purpose. Learn how to properly label financial time series. There are chapters in books written just about that.