Guided project - Predicting stock market price will not make you rich

Hi all,
did this project locally within a jupyter notebook as the console environment seemed to be very laggy for me (changes in the editor were not properly saved when running the script via the terminal).

I make a statement, that the prediction does not work at all. Change my mind. Maybe I am overseeing something or did an error?

Thanks and best wishes
Gunnar

https://app.dataquest.io/c/11/m/65/guided-project%3A-predicting-the-stock-market/7/next-steps

predict-charts.ipynb (263.7 KB)

Click here to view the jupyter notebook file in a new tab

1 Like

If you zoom in on the plot of predicted price vs real price then all those great results make sense:


the model is just trailing the real price, it’s not actually predicting, but the way the lesson is set up you’re supposed to be happy with it because rmse is very small.
I’ve tried training a different model that would just classify the next day as a “price going up” or “price going down” day (a real trader would prefer that info) and yeah… no wonders there couldn’t get to 53% accuracy with single model, stacking a few together almost reached… 54%

Thanks for the confirmation!
I also tried something similar (price up or price down) and i think the most simple linear regression model only reached ~50.5% correct.
I was only wondering when the course screen said something like "now we successfully implemented a stock prediction :smiley: