456 points by predictivemodelinginc 2 years ago flag hide 13 comments
john_doe 2 years ago next
Fascinating topic! I've been researching ML for stock trends too. Any recommendation on the best type of dataset to use for training?
ai_expert 2 years ago next
Hey @john_doe, I suggest using price data, volume data, and other relevant data points like economic indicators and company earnings reports to train ML models.
another_user 2 years ago prev next
I wrote a blog post about this specific topic a while ago: <https://medium.com/@another_user/dataset-selection-for-stock-trends-machine-learning-9d56a1b>.
random_dev 2 years ago prev next
Is there a difference between using traditional ML algorithms and deep learning when predicting stock market trends?
ml_guru 2 years ago next
Both traditional ML and deep learning have their advantages and disadvantages. Deep learning can model more complex relationships, but requires more data and computational resources.
curious_engineer 2 years ago next
I'd love to read more about your experience using an LSTM-based RNN for stock market predictions. Do you have a link to a repository or write-up?
quant_analyst 2 years ago prev next
We've done a thorough evaluation and found gradient-boosting to offer the best compromise.
tesla_investor 2 years ago prev next
I've seen some brilliant success with an LSTM based RNN. Give it a try.
student_ml 2 years ago prev next
What are the necessary preprocessing steps before training a model?
hands_on_ml 2 years ago next
You'll want to focus on feature scaling, normalization, and handling missing values in your dataset. Additionally, consider applying relevant data transformations and feature engineering.
ml_enthusiast 2 years ago prev next
I've tried my hand at predicting stock market trends and found the results to be hit or miss. What are recommended ways to evaluate model performance?
stats_prof 2 years ago next
Consider using metrics like mean squared error (MSE) for regression tasks, or accuracy and F1 score for classification tasks. Don't forget to evaluate your models against appropriate benchmarks too!
ml_enthusiast 2 years ago next
Thanks for the advice! I wasn't aware of the F1 score before. I'll take that into consideration for my next project!