This was a 2 to 4 weeks project. Trading on cryptocurrency markets. I never used it with real money because I wasn't able to get it to be consistently profitable. I probably have some code lying around somewhere on one of my servers. It was a pretty basic/rough solution though but it had basic machine learning. I was using Node.js with a simple open source Artificial Neural Network (ANN) library - I think it was Brain.js: https://www.npmjs.com/package/brain.js?activeTab=readme

I think maybe if you use technical analysis (price and volume), you may want to combine it with some other data too. For example, for crypto, I was thinking to combine it with blockchain analysis - That way, if someone moves a lot of cryptocurrency tokens to an exchange wallet, that can mean that a big selloff (and price drop) is about to happen. Also, maybe feeding a real-time news feed with sentiment analysis into the algorithm would help too.

I gave up on this because it would be time consuming to collect all that additional data and I was just trying things out looking for a low-hanging fruit. There's no guarantee it would work even if I got the data.

That said, I think my setup was pretty good; I just collected some historical price candlestick data via some API calls to the exchange/trading platform; then I would split the result set in half, use the first half for training the ANN and the second half for testing it. Then I would try with different settings, increase the training time, etc...