User transformation

The user features for a specific user tweet are first transformed into more useful features. Here is example of the pre-transformed user features. Note: Each feature has an associated stock specific feature (Ex. accuracyandaccuracy_s )

Below are the new features that are generated.

num_tweets: number of total predictions

accuracy: ratio of correct predictions to total predictions

return: theoretical return based on past predictions

return = total_return - return_s

Subtract stock specific return so that total return isn't skewed by return_s

return_w: theoretical return weighted by time of predictions

return_log: theoretical return weighted by number of predictions