Stock Trading AI: Using Alpaca & Stable Baselines for Reinforcement Learning Investing

Hey, thanks for clicking on the video.

I talk about coding, python, technology, education, data science, deep learning, etc. If you enjoyed the video please like or subscribe. It is one of the best ways to let YouTube share similar content to you and others interested in this topic.Many thanks

CHAPTERS
Alpaca - 01:06
Even More Custom Indicators - 05:35
Stable Training Policy - 10:13
Results - 12:58
Outro -14:10

My goal is to create a community of like-minded people for a mastermind group where we can help each other succeed, so browse around and let me know what you think. Cheers!

Keyword for the algorithm: A2C model with historical stock data for AAPL and stablebaseslines3 packages create better results in our reinforcement learning agent neural network deep learning machine learning in finance data science python project

FOLLOW ME ON LINKEDIN:

[SM-yhmj]

Great video, thank you very much for sharing your expertise

[beyondboundaries]

Awesome video, cannot wait for you to investigate FinRL!

[ay]

Cool, this is interesting. It seems that the input to your model is the historical prices, vol or stuff that can get calculated using these. Note that you should check beforehand whether any of these have any predictive power or not. From my experience in a lot of cases for example lagged returns could be completely independent from the future's return, so including it or any function of it as an input might not help at all. I think in order to fairly assess the model one needs to pay attention that what we are feeding to that model. At this point it seems that you are training a sophisticated model to detect something that doesn't even exist. Usually you will need to come up with signals that have predictive power and feed those to the model.

Another thing, is the data format open, close, high, low. I have no idea why this is so prevalent but it is completely useless in my opinion. You will probably need at minimum bid/ ask prices and ideally orderbook data. If you get an orderbook data, it should be possible to train agents to do high frequency trading. Which is more predictable and doesn't require complicated signals.

[davekelly]

Great video. Unfortunately your results are similar to my experiences using OpenAI-based RL for Forex trading, which I find surprising. This implies there is no correlation between market history/technical indicators and predicting future market movements, or there is something wrong with the RL algorithm. Did you work out why it never took any Short trades? Interested to see how you get on with FinRL

[brianferrell]

Hey curious if you have looked into the finrl package for stuff like this. I am just getting started with it

[vungouc]

_process_data = singals
Sorry Sir where do you use signal function in your code? I see you defined but not use?

[hom]

This is brilliant, is the source code available on github?

[mambo]

all the RL trading videos I seen, don't not make any profit