How to Train AI to Day Trade Crypto with FinRL and Python
Вставка
- Опубліковано 15 жов 2024
- Hey, thanks for clicking on the video.
JOIN NEWSLETTER: witty-motivato...
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 UA-cam share similar content to you and others interested in this topic.Many thanks
CHAPTERS
Getting Data - 1:00
Make Environment - 6:25
Train & Test Agents - 11:10
Back Testing - 13:00
Results & Things I'd do differently - 16:54
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: data science finance deep learning finrl crypto currency a2c deep reinforcement learning in finance drl stock trading day trading python coding ai artificial intelligence trades crypto currency
FOLLOW ME ON LINKEDIN:
/ everett-minshall
YOU GIGA CHAD! Thank you!
Please include the notebook download link as well. thank you very much
I'm sorry guys, I really am trying to remember to include the code. I just forget to commit to a repository or create a colab version.
Very interesting and we can always improve the TA so far so good 👍
Thanks for the short and crisp working explanation, truely appreciate it! Can you pls let me know how can I make predictions on hourly/daily basis on live (dummy) data for testing with Alpaca. The test/trade data you used worked on series of pre-available data in a df, how will that convert to output for the last updated timestamp? I'm guessing it should be something similar to step, action, reward we get from a RL agent.
Hello, might be a dumb question, but how do you think we could use FinRL for real trading? Is it fully automatic, or can I set it to just alert me for buy/sell decisions? I'm more into the idea of getting notifications and handling the trades myself.
Not a dumb question at all, I get a lot of questions like this. I haven't really thought of a way to deploy a model to real world production because a) I've never seen a model perform well enough to actually use and I discourage people from trying honestly. b) I'm just not that experienced yet to write the code to do that. I would probably simulate a data stream similar to the crypto market and pass the data through the model that had functions to push buy/sell orders then trigger a notification to my phone whenever it made a trade decision.
awesome video, would love to play with the colab
Oh right on thanks for reminding me to include it
Something seems off about the MVO when compared to the A2C model. Why is the start value 5.56e6, surely this cannot be the portfolio value? Why is it compared to the portfolio value of the RL agent?
Admittedly it isn't the best benchmark, but MVO is a popular enough portfolio strategy that I think it serves to point out how trading with RL isn't as great of solution out of the box as it seems compared to other methods. If I remember correctly, the code for MVO is taking the weights across the entire dataset from training to validation so it's starting higher because it's only a snapshot of the end. This was maybe a mistake on my part to set it up this way.
Is anything of this transferable to getting it to trade forex instead?
Yeah, the difficult part is making sure the environment the agent is training in simulates the actual market conditions, so I either have to build a forex environment or modify some code I find online which takes time. I'll try to make a video on the concept in the near future.
@eminsh oh brilliant! I'm trying to study myself but still think I'm a beginner. Was using blqckbox to help guide me on creating a learnong model with mql. Didn't work too well though 😂
Was looking for tutorials on that everywhere but couldn't find much. Looking forward to it if you do make one anyway. Subscribed!
Please can you do a tutorial for forex using user imported data sets....
I will look for a forex env. What do you mean by user imported data? like data from an api or csv?
@@eminsh yeah yeah data from a csv...if you are interested we can have a joint project together....i have some ideas i think we should try...how can i reach you
how do you resolve dependencies? Poetry can help with that.
I must be doing something wrong because I couldn't get it to work despite using poetry. Still trying though when I have time and motivation.
I am sort of making a real trading system utilizing some of the FinRL stuffs. One problem i had is that it act weird at the beginning of each trading day. I guess there is something need to be improved as stock trading is different than playing a game continuously but haven't sorted out yet. Maybe someone here has some ideas to share
Yeah I've been thinking about this too. In a game like chess the environment doesn't change too dramatically unless the agent takes an action; whereas the state space of the markets change constantly whether the agent acts or not. I think it's similar to self driving.
How would you connect an RL agent to MT5?
I m also still searching about this issue. Please update here if u manage to figure it out...i will also do the same.
wait you are from Zim wanguuu 😂😂😂
😂😂😂We made it here mdhara kkkkk @@JonaKatsande
I'm not sure, let me research MT5
@@eminsh Mt5 is for forex trading, do some tutorials for currency pairs or metals in forex.