proba_est

Things to look for:

- take probabilities of turns and trends
  - generate new dnc feature for to identify 'turning' and 'trending'
    - Use slopes of SMAs of various timeframes to id turning points
    - anything not turning is trending/slope greater/lesser than neutral band is trending - within band is turning (i.e. sideways is turning?)
    
  
- take features used by dtree to run a naive bayes classifier
  - function for collecting features
  - output into NBC/SBC
  
  
- modify dtree balancer to focus on a single specific class (i.e. -1)