80 points by fintech_guru 6 months ago flag hide 12 comments
john_doe 6 months ago next
Fascinating topic! I've been curious about how machine learning can be used in this way. Hope to see more insights from this thread!
ajith 6 months ago next
Definitely worth looking into! I wonder if anyone here has tried applying machine learning to stocks beyond just time-series predictions and academic research?
quant_surgeon 6 months ago next
Yes, I agree, academic research is the tip of the iceberg. I've been using complex ML models in my high-frequency trading strategies. Happy to share my experiences if that helps.
jane 6 months ago next
Hi @quant_surgeon, I'm very curious to find out about your HFT strategies. Do you use any specific libraries or types of ML models? Care to elaborate a little more?
quant_surgeon 6 months ago next
@jane In short, I use SOHCAHTOA, with a mix of Gaussian processes, random forests, and LSTM networks. I'll leave the acronym's breakdown in my next reply!
jane 6 months ago next
Thanks for the response @quant_surgeon! I'm interested to know more about how SOHCAHTOA applies to stock prediction in your ML strategies.
ml_scholar 6 months ago prev next
There's quite a bit of research being done on stock prediction via ML, but I agree with john_doe, let's discuss it further here. Check out this new paper on using LSTMs...
stocks_newbie 6 months ago next
Thank you for sharing the paper, ml_scholar. It looks very complex though, and I have trouble understanding the math behind it. Any simplified resources to learn about LSTMs?
educated_coder 6 months ago next
There's a great interactive tutorial on KDnuggets about LSTMs: (link). You should explore YouTube videos as well, specifically those made by Siraj Raval.
stocks_newbie 6 months ago next
@educated_coder the YouTube videos were helpful for getting a better understanding. What about visualization tools for LSTM predictions? I couldn't find any examples online.
educated_coder 6 months ago next
Stocks_newbie, there are a variety of libraries that help with XGBoost and TensorFlow-based ML models that can generate visualizations. Plotly, Bokeh, and R's ggplot2 are some great examples.
stocks_newbie 6 months ago next
@educated_coder, thanks for the visualization tip! Still having trouble grasping LSTMs, but finding more online resources to slowly learn the math behind it.