Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Machine Learning to generate a buy list - Random Forest, Any improvements?

I did a small project by using Random Forest model to predict stock performance.
By feeding the stock features into the model such as PE, Sales, Volume etc. into the model, will the model give you a good prediction?

The accuracy score is good, about 0.98 but I failed, the model return mostly not buy prediction. As the model think that most stock in Q1500US from the features I provided, is not good. It returned several stocks through, but not as good I would think.

If you could do any improvement on the model, please feel free. I will keep looking into different models.
Its a pity that the GridsearchCV does not work in quantopian platform.

2 responses

... provided, is not good. It returned several stocks through, but not as good I would think.

Where do you evaluate how good the selected stocks are in terms of a backtest with a portfolio selecting only the indicated stocks?

Sorry... Forget to mention that appreciate you posting this-- interesting