103 points by quant_finance 6 months ago flag hide 13 comments
johnsmith 6 months ago next
Fascinating topic. I've been experimenting with machine learning techniques for stock market prediction as well. I mainly use LSTM networks with good results.
alice25 6 months ago next
LSTM networks eh? I tried Random Forests, and they often surprised me in their accuracy. Though, tuning the model did take some time.
mlfanatic 6 months ago prev next
Consider using a combination of technical indicators and price data as features for your machine learning model. This should yield improved results.
deeplearningjoe 6 months ago prev next
Are any of you trying to predict specific stock prices or solely market trends?
bigdataisgod 6 months ago next
I target specific stock prices, as most trend-based models I tested lack enough accuracy to make them useful.
predictionbro 6 months ago next
To predict specific stock prices, I find that combining different models can boost performance significantly. Ensemble models like XGBoost or Stochastic Gradient Boosting (SGB) work well.
algorithmgorilla 6 months ago prev next
I predict market trends as using separate models for a wide range of stocks consumes a lot of time and resources.
mlstocks 6 months ago prev next
If anyone is interested, I've compiled a list of useful repositories and tutorials related to machine learning for stock market predictions.
johndee 6 months ago next
That'd be super helpful, especially for people just starting to learn about these applications. Thanks for sharing!
victor 6 months ago next
@mlstocks, thanks for compiling your list. I've implemented a few techniques discussed there, and have already seen small improvements in my model's accuracy.
algoqueen 6 months ago prev next
Are there any Quantitative Finance types here familiar with the theory behind Machine Learning techniques applied to financial prediction? I could certainly use some insights.
cointrader 6 months ago prev next
Does anyone have experience with applying machine learning techniques to crypto markets? Curious if the same approaches work.
nostradamust 6 months ago prev next
There are numerous challenges when using machine learning for stock predictions, such as noisy data, the need for massive amounts of historical data, and dealing with extreme market situations. Tread cautiously and ensure that your models can handle such scenarios.