156 points by tradetiger 1 year ago flag hide 13 comments
john_tech 1 year ago next
Fascinating project! Real-time stock trading bots can be real game changers. Anyone tried implementing this with ML algorithms like LSTM or GRU?
algo_queen 1 year ago next
I've used LSTMs for similar projects, and they can be quite accurate. However, tuning the model for optimized performance is key.
trader_alex 1 year ago next
Indeed, hyperparameter optimization is crucial for LSTMs. Would you mind sharing some useful resources to learn more about them, algo_queen?
algo_queen 1 year ago next
Of course, trader_alex! This resource has worked wonders for me: https://machinelearningmastery.com/hyperparameter-tuning-time-series-forecasting/
ml_ninja 1 year ago prev next
This is impressive! May we know the programming language you used to build this?
code_guru 1 year ago next
Python was used for this project, as it's quite popular in the data analysis community for its extensive libraries and user-friendly syntax.
code_guru 1 year ago next
Apart from the usual normalization and missing value imputation, we employed a few additional techniques, such as window-sliding and feature scaling, quant_dave.
code_guru 1 year ago next
We carefully trained and tested the model on separate datasets to make sure there were no data leaks. Cross-validation was also employed for further validation, market_wizard.
code_guru 1 year ago next
We implemented several risk management strategies, including proper position sizing, stop-loss orders, and periodic backtesting to monitor model performance and tweak as necessary, data_sage.
quant_dave 1 year ago prev next
Curious to understand what kind of data preprocessing was involved in this project, code_guru?
market_wizard 1 year ago next
Data leaks can be a real challenge when building prediction algorithms. How did you address those, code_guru?
data_sage 1 year ago prev next
Making predictions in real-time for stock trading couldbe incredibly risky. How do you plan to manage risk and avoid significant losses?
risk_manager 1 year ago next
@data_sage, we usually apply position sizing techniques like Kelly criterion, fixed fractional, or fixed ratio to manage risks associated with real-time trading. Also, implementing a trailing stop or a circuit breaker can help minimize losses aswell.