We can also build ensembles with models using different horizons. Here are the results of a 6 to 18 MSE/MLE horizon ensemble (12 models in total), using a majority vote (take signal if most agree) or an absolute vote (only take signal if all models ...

Machine Learning with algoTraderJo

Changing the MSE/MLE horizon has a very important effect: image

Machine Learning with algoTraderJo

Now some results for the USDJPY. This test uses an SL of 50% of the 20-ATR (same as before), uses 80 examples for model building on each signal hour, uses 6 returns as inputs and trades on hour 1 GMT +1/+2. The MSE/MLE horizon is 12 bars as well. As ...

Machine Learning with algoTraderJo

I do not cap the predictions in any manner but I do not get negative predictions within my model (probably by chance, I do see how it can happen). However this should not be problematic if you are doing a simple comparison to generate signals, ...

Machine Learning with algoTraderJo

For me it never happens because the MSE and MLE are calculated from the current open. If all values within the next 12 bars are above the current open then the MSE is 0, in no case can it be negative.

Machine Learning with algoTraderJo

I am using 7 pips as the spread. As you point out the spread on this pair can be bigger, so I used 7 as a compromise of the historical average. I also did a small simulation using tick data (2007-2014), results are very similar. Now to answer your ...

Machine Learning with algoTraderJo

Here is the code, note that F4 functions are used for getting price values, etc. RegressionDataset regression_i_simpleReturn_o_mfemaeDifference(int period, int barsUsed, int hourFilter, int successBarThreshold){ Data<RealVector> ...

Machine Learning with algoTraderJo

Couldn't have said it better

Machine Learning with algoTraderJo

Time is also an incredibly critical factor... Try to train/predict at the wrong hour and you get much worse results (remember that examples are also built using previous days at the same hour). It is no coincidence that the best results for this ...

Machine Learning with algoTraderJo

The observation about the 12 hour relationship is interesting indeed Is the 12 hour target horizon and the best 12 input result a coincidence or something of more substance? How would you test this to find out? Looking forward to your questions ...

Machine Learning with algoTraderJo

I use the Shark machine learning library which is implemented as part of the F4 framework. I use the linear regression shark class as explained here. The Shark library is completely free and open source so you can implement this within your project ...

Machine Learning with algoTraderJo

It is an additional bar return. The 8 input curve uses 8 bar returns on each example, the 9 input curve uses 9 bar returns, etc. The difference between the 8 and 9 input curves is simply the use of one additional bar return for the building of ...

Machine Learning with algoTraderJo

Going back to the EUR/JPY machine learning strategy I just posted, here is what happens when you change the number of inputs: image What does this tell you? Any comments?

Machine Learning with algoTraderJo

Yes, Daniel and the Asirikuy community in general are much more focused on the production of price action based systems. I am working with Daniel on some machine learning setups and this is mainly the reason why I resumed posting here (Daniel ...

Machine Learning with algoTraderJo

Sorry, had a couple of years missing from the last backtest I posted here. This is the full version (1991-2015). The last year has been significantly profitable for this system. Remember that the cumulative return axis is logarithmic. image

Machine Learning with algoTraderJo

Thanks for the warm welcome

Machine Learning with algoTraderJo

Long time no see folks! Life happens, we get busy, but now I am back! I want to start posting again on this thread to get the ball rolling again on machine learning in FX trading. To start things fresh I want to start sharing with you a machine ...

Machine Learning with algoTraderJo

Things have become a bit quiet now , perhaps you are missing some information to figure out what is happening here. I will now show you another, more radical, ensemble example with some additional modifications that may shine some light ...

Machine Learning with algoTraderJo

Much better! However you are still thinking as if an ensemble was a portfolio (thinking that the returns add to each other and so opposite returns would yield better results). The success of an ensemble cannot be predicted from the return series of ...

Machine Learning with algoTraderJo

Please read the thread, it explains in detail the input/targets I am using for the training of machine learning models.

Machine Learning with algoTraderJo