Looking for Advice on Using Machine Learning in Predictive Trading Techniques

Hello Everyone,

Since I’m not too familiar with machine learning or its use in trading, I’m looking for advice from more seasoned members from this community.

Despite my expertise in finance and my limited Python coding skills, I’m looking forward to the challenge of incorporating machine learning methods into predictive trading techniques. Specifically, I’m interested in creating models that, using past data and other pertinent variables, can forecast changes in stock prices.

These are some inquiries I have:

Beginning with Data: How should historical market data be gathered and prepared for machine learning designs? Exist any instruments or data sources that are suggested to help speed up this process? :thinking:

Selecting the Correct Model: Which procedures would you suggest a beginner in machine learning start with for someone with an intermediate degree of experience? :thinking: Though I’m not sure which would work best for trading, I’ve heard of neural networks, decision trees, and linear regression.

Feature Engineering: Which essential characteristics or indicators are most helpful for predictive trading models? :thinking: Although I’ve read about RSI, MACD, and moving averages, I’m still seeking more information on feature selection.

Assessment and Retesting: What are the best ways to assess my models’ performance? Would you suggest any particular metrics or retrospective testing frameworks to make sure the model is trustworthy and stable? :thinking:

Resources and Educational Materials: Is there an extensive guide to automated learning for trading that you would recommend in the form of books, online classes, or tutorials? :thinking: Resources that address both theoretical ideas and real-world application are what I’m searching for.

I checked this :point_right: https://medium.com/analytics-vidhya/how-im-using-machine-learning-to-trade-in-the-stock-marketsalesforce

Thank you :pray: in advance for your help and support.


I suggest to read Stefan’s book while doing the proposed examples and get some understanding of the math involved.

Technicals only in general do not work as well as with fundamental factors, there are some discussions on the book around fama factors that’s interesting.

I’ve seen some recent research of an approach with cci, bb & rsi over dailies on SPY(RTH only) I’m checking it over 30min on NQ and is also promising, have yet to check how it performs short in 2024.

Linear classification with logistics regression is what I plan to use to classify factors (I’m doing data mining now). Check chapter 7 on the book.

As of ML, seems that deep learning can be the way to go (not arrived there yet).