let me try also.
To aroonraj:
Can this system be made into an Ea?
hi,aroonarj,
Your system is very good,but I have a question.
Is the system needed to update offten?And how to train the NN again?
Thank you!
i guess he done making ea. doesnt need us.
Hi nabil114,
I wish I could one day. Due to various reasons I could not contribute actively in this thread and on the project. I am still in search of a good model. My experience says its difficult (I wont say impossible) to approximate a function for forex markets.
I am currently working on classification models and am using SVM (Support Vector Machines). Also I no more use JNNS for training due to precision limitations. I am still interested to actively reply in this thread or a new thread on NN based EA development.
Best Wishes,
Arun
USD/JPY
7 - 8 gmt + 1 high or low.
- stoploss is 30 pips
- take profit is 3 pips
- dont take trades at round number to be safe place trade at 94 for short and 04 for long.
- make sure of near support or resistance levels close (at least a pip up or down) to your take profits or order entry then place order entry below or above
- order entry dies at 14:30 gmt +1
- wait for 10 minutes to start pending order reason because in few minutes prices might follow trend of previous prices might create false breaks. this would happen because buyer are drying up in the next lag.
try this.
it my method.
<script src=“http://meetpips.com/profiles/nabil114/widgets/script.js” type=“text/javascript”></script>
working well for a while now.
It seems that I have arrived late at the party. The thread seems to be dead or dying, but it is of great interest to me, so I post anyway in hope someone is still out there.
It seems that one difficulty with this approach to a net for FX is that one must choose an indicator as a target. I have been experimenting with NN’s for FX for several years, and have been able to train many nets to predict an EMA for example, with very small mse. The problem always seems to be that all the error occurs at the pivot points, just when you want accuracy. I’ve never been sure that the net was any better than a simple linear prediction, which is also very good except at the turns.
There is another approach to be found in a recent article on the MQL site “Using Neural Networks In MetaTrader”. The author has managed to integrate FANN into MT4, so that one can train a net based on how well the associated EA performs. This avoids several problems discussed above: 1. One does not train for mse, but instead for signals that give positive return. 2. One does not have to set arbitrary thresholds for the output signal, but can find the best threshold by genetic optimization over the training set.
This latter approach has it’s own drawbacks. First and worst, the training must be done manually. This takes a great deal of time. It can be overcome by writing a windows script (in autoit e.g.) that does the training, but it will still take lots of time. Second, there is no possibility to use a validation set to decide when to stop… this must be done manually. Third, the available networks are restricted to the set in FANN, so that other nets like recursive Laguerre, that are better suited to time series, cannot be used.
Finally, if anyone is still interested, it would be great if a competent programmer ( I am incompetent) would integrate a more powerful Net training program with MT4 and then somehow include a script, or other mechanism to allow training and validation.
Hi,
I truly agree with your opinion. Using NN for forex is an art by itself. Even I have been experimenting since a couple of years the applications of NN in forex trading. Although I have abondoned this threada and also many other projects due to various reasons, i still, in private, keep trying new models. I tried almost all kinds of NN approaches
- Function approximation
- Classification
- Clustering methods
Recently I started working on currency strength based forecasting of trading signals. If I get some good results, I would definitely share it in this forum. I know many people are interested in NN and they keep posting me asking questions how they can use NeuroTrend, my experience says its not easy to apply NN unless and untill you define a proper problem space. The type of network, type of method used is secondary.
[B]In defining the trading strategy model lies the whole solution for development of successful trading system and not in Neural Networks.[/B]
Best Wishes,
:):)Arun:)
Hi,
I’m no programmer at all, but I wonder if this good indicator can be used for other currency pair (or other time frame) than Euro/USD M15.
There are 2 files going with this indicator, NeuroTrend_Include.mqh and neurotrend.net.
What/which file should be changed (trained?). And how can I do that?
What is the purpose of NeuralTrainingsetCreator.mq4 , and how can I use it?
hi,
I’m very new to the whole fiels of stocks, forex and NN. I’ve stumbled across NN during my regular job and figured out via google that it is possible to make more or less accurate stock predictions. So, that was my idea: make a good stocks model on a daily (a few days max.) basis.
Because I want to test the model as often as possible forex would be ideal, because one can make a lot of trades to force out variance. Thats why I end up in this forum.
But thats not what I wanted to ask… I’m a little confused about the idea behind: feeding in data from the (say) usdeur chart, and predict the same. Does the market behave like this? I mean one cannot predict the future on a single chart? The whole markets dictate usdeur and whatever. So my idea for the stockpredictor was, feed in a few input stocks, maybe start with sth between 5 and 25. As target variable I use just a simple “up”, “down” or “buy” / “sell”. The stock I wanted to predict is NOT in the input files (aka the learning of the network. Then I can calculate a correlation between input-output and use other stocks, use other models etc.
For the prediction I just give the model the values for say DOW, Goldrpice, oil … and the model says “buy”, 60% accuracy. (actuall its not THAT easy, but thats the basic idea behind).
But I will try forex/the thoughts given in this thread with some models. Mabe we can even start a discussion here again.
best regards,
Stever
Hi Stever,
The initial idea that I shared in this thread was just for discussion and as a start to find better models. I agree that one need to consider as inputs other forces too for prediction. Many forex brokers do not have option to trade other comodities so its not possible to get the data for other commodities.
I currently use data from different currency pairs (inter-market) to predict the strength of EURUSD. Its still in research so I cannot mention any results, but I agree that predicting based on the history data of a single currency pair and its derived indicators is not very reliable. Your winning % of 60 is good as a start. if you can achieve the same with your model in forex, you can create reliable trading strategies.
Also prediction on daily basis will show good winning ratio in training and testing NN, but poor results practically on accounts with small amounts. I am not a full-time trader and am not an investor so I use smaller timeframes (M1-M15) for forecating signals. You can think my trading style as a NN scalper.
I am open for new ideas and discussion either in this thread or in a new thread on building trading systems based on NN.
Hello
I am studying NN too. Once I had reached same colclusion as you Arrone (i.e. predicting based on the history data of a single currency pair and its derived indicators is not very reliable.)
I still have not achieved reliable accuracy using NN especially using self data.
But including myself, some people can predict over 70% using only self data as input using some statistical methods in university text book (for example, None Neural Network method).
From this fact, my believe is that it should be possible to predict over 70% of accuracy using neural network too only using self data or using derived data from self data.
I think the problem comes from neural network itself.
and My view about NN is as followings:
Firstly, it has too much parameters while we only have limited number of inputs for training.
Secondly, it is too slow.
Thirdly, it often gets to wrong path to the solution while training (if you use many different training methods including genetic algorithm, steepest descent, conjugate gradient and so on you will see and feel this.)
Anyway, just written few comments as I think I am on where you are too in terms of NN wise.
good luck !!
Didn’t anyone notice that NeuroTrend is seriously flawed ??
I made a simple EA, trading on NT signals, and backtested it.
The result was amazing… profits went trough the roof in no time.
There was obviously a problem, and after a quick glance at the NeuroTrend code I found that the internal inputs 15 and 16 is taken from 1H EMAs which
will be looking atleast 45 minutes into the future. Works great on backtest
where 1H bar0 is always finished and available, but obviously fails miserably on
forward test. I did forward test it and as expected the results were nothing
near what the backtest showed.
So it looks to me that NT in its current form has no future…
Neuro Trend indicators are universal, working with any market and any time frame. This includes Forex, Stocks, Futures, Bonds, Equities, Commodities. Neuro Trend indicators are user friendly, allowing the trader the ability to adjust the variable inputs for their own style of trading. We`ve used the most advanced technology available in the designing and developing of the Neuro Trend indicators.These indicators are one of the best sets of indicators ever created by our developers. These indicators are easy to use, even though they are programmed with the latest and most Complex Logic Algorithms available today . Investors and Traders can enjoy a greater potential for profit, including minimizing your risk. Neuro Trend indicators work in real time or position trading. Neuro Trend will work in any time frame and any market, including: Tick, Minute, Hourly, Daily, Weekly, Monthly or Yearly. The objective of every trader is to make a profit. That is what we had in mind when we developed this software program.The NeuroTrend indicator generate the buy and sell signals. It clearly shows the overall trend. It help maximize your profits and minimize your losses.
So what are you talking about here?
I was talking about aroonraj’s “NeuroTrend”.
I tried this for a couple of days now and it’s pretty amazing. Only time will tell now.
I am confused by your comment. Inputs 15 and 16 refer to the previous H1 bars since shift is 1 and 2. This does not seem to look forward. All the documentation on the strategy tester used for backtest state that the ST simulates ticks one at a time, and will not allow your EA to look into the future. It is not unusual for an EA to backtest well and then fail in forward test. There are many reasons for this, but it is truly hard to get the ST to look forward… you must fool it by using an external file with the output of the net, previously computed, but shifted forward, or something like that. Perhaps the net was trained on old data and needs updating for forward testing. Or perhaps the backtest was on the data for which the net was trained… that’s just curve fitting.
On another subject, has anyone tried to use the volume and range of the current (shift=1) bar as an input? There are many who claim that Volume spread analysis (VSA) really Volume Range analysis, results in profitable trading. They claim that patterns in the VSA data can be read and used to predict one or two bars ahead. I have made experiments using a K-nearest neighbor classifier and have found that indeed there are patterns to recognize. One can predict the direction of the next bar with something like a 63% accuracy, where the accuracy is measured as the number of positive or negative predictions that are correct. It misses many, by calling them flat, and though these are errors, and would make the overall error rate about 50%, they aren’t important if you don’t use tem to trade… So far I have not been able to use the prediction in an EA that returns acceptable profits, but I thought that some VSA inputs might improve neurotrend. If someone is interested, I will attach an indicator that calculates four VSA features that I have found to be useful.
No the EA gave HUGHE profits on any pair, month, or year, and without any form of re-training.
The price of 1H bar0 is its closing price. If you access this bar from a lower TF eg 1M you’ll have that price available for a full 59 minutes before that
1H bar actually closed… iaw looking 59minutes into the future.
You can monitor that bar0 price in backtests and see that it is actually stable, but on foreward test the price moves with the ticks…
Hi, MadCow,
What is VSA? is it derived from volume (cltr-L) in MT4? I would like to study your VSA indicator please.