Discussing the article: "MQL5 Wizard Techniques you should know (Part 96): Using Wavelet Thresholding and LSTM Network in a Custom Money Management Class"

[MetaQuotes]

Check out the new article: MQL5 Wizard Techniques you should know (Part 96): Using Wavelet Thresholding and LSTM Network in a Custom Money Management Class.

In this article we consider a custom MQL5 Wizard class that processes Money Management. Our custom class is labelled ‘CMoneyWaveletLSTM’, and is developed by combining the Wavelet Thresholding algorithm with an LSTM network. As has been the case throughout these series, the developed model is testable with MQL5 Wizard-Assembled Expert Advisors that can be tuned with different trailing stops and entry Signals classes. We maintain our entry Signal, as in past articles as the built-in 'Envelopes' class and the RSI class.

We continue this series on different ideas and trade setups that can be explored and tested thanks to the MQL5 Wizard. In the last article where we looked into money management, in our custom class, we married the Suffix Automation Algorithm with an Autoencoder network. The model we had then could be the suitable tool for traders that like exploiting consolidating markets and where the historical "Price DNA" provides some repetitive patterns. It appeared to thrive in these conditions, provided we had a rhythm. This could then beg the question, what happens when the regime changes. For some traders that test volatile markets such as intraday forex or crypto, repeatable and dependable patterns can be rare. The primary obstacle here often seems that the market is static-erratic with high-frequency noise often disguising as prevalent trends.

For this article, we thus move from pattern recognition to signal processing. Instead of looking out for macro repeatable patterns we now attempt to come up with a surgical tool that is best suited for noisy, momentum-driven environments. We build this tool as a custom money management class that merges the algorithm Wavelet-Thresholding (adopted to de-noise log returns) and a Long Short-Term Memory (LSTM) network. We are sticking to the dual-engine approach of our recent articles as we continue with the theme of introducing more adaptability in scaling position size as we strip away static and lagging signals.

Author: Stephen Njuki