PNN for metatrader using Parzen window classification version 2 with kernel smoothing of inputs

This is the PNN using Parzen window classification version 2 with kernel smoothing of inputs. You need to have the PFE indicator.

Here you will file the standard PNN called version 2. And the PNN with money management and kernel smoothing of inputs.

Nevertheless this is a beta version, because I am not sure that the integration of the kernel smoothing with the PNN code is correctly executed.

So this is what we have for now. Anyway the idea is interesting.

In the normal PNN Eric we have as input the difference between:


b[bar]= Close[bar]-Open[bar+AmoutOfForecastBars-1])

The idea was to replace this piece of code with something else and to see how it would affect the EA.

So I replaced this by the kernel smoothing of PFE.


b[bar] = kernel()

So this is the idea. However the number of bars of parameters is still an important parameter because it affects the initial period for training:


TrainingStartTime = Time[0] + Period()*60*AmoutOfForecastBars

Of course we can replace this piece of code by Numbars_for_Training as a parameter, but I chose not to touch this for the moment.


TrainingStartTime = Time[0] + Period()*Numbars_for_Training

Any wise look in the code would be welcome as there is not many native mql code for neural implementations.

DOWNLOAD from here

Human brain and parallel signal patterns

As arryex from the Trade2Wind I also have been inspired from the this article. It is about the Neural net application from Trader Tek.

They claim that their Pipelined Recurrent Neural Network performs better than the the human brain recognizing sequential signal patterns such as stock prices.

Of course the beginning of 2000 that was the explositon of the idea that Artificial Learning algorythms are the holy grail. There are many commercial and free softwares available. Of course the obvious idea is to have a model for yourself. By now this is accessible to everybody, you can download and test arryex things or you cand download and test the Back Propagation Neural Networ (BPNN) with a whole library of inputs (SSA Caterpilar, Hodrick- Preskott, Jurik type of digital filtering, Volume Weighted Average Price VWAP, etc. etc.). However you will not get any edge.

The think is and there is the second part of the article.


whereas the human brain is better at recognizing "parallel signal patterns"—the human eye can easily tell a picture of a tiger from a picture of a cat.

And that I think is important. And that is why I try to make use of human pattern recognition tools that would be able to add an edge in Neural Net applications for trading purpouses. The first approach is the Elliotware but I also try another approaches as the Market State Analysis.

Using neural net models newbies should be very carefull. It may be boring but I have to say. Look at your fundamental calendar, do not be surprised, the market orders accumulation is also not accounted in the pattersn on which is trained the neural net. And there are so many things that are not in the model, so do not expect much from it.

It performs well when it is trained on the current market activity and you expect a microstructure event (I do not expect any neural net to predict a major top and bottom) provided nothing serious is happening right now. Just make sure of that.

I really liked too much a phrase from quantin123 it is not necessarily a joke:


Always keep in mind that NN is just a dumb calculator.