Machine Learning Bot Architecture MQL5

 

Hi

What architecture would you recommend for a machine learning bot in MQL5?

My opinion based on my experience with python only and Oanda REST API:

1. Training independently (Python etc). Save model as REST endpoint (in my case i was saving it as a pickle file)

2. Trading logic in MQL5 and prediction via internet (in my case i was loading the pickle file)

I assume that any machine learning bot needs a training part (runs periodically) and a trading logic (runs continuously during market open times)

Is there any way I can do training and prediction in MQL5?

Is there a library that can be used like eg scikit-learn?