Prototype draft

[teobucci]

Design

Inputs:

• data in CSV format
• text input for the name of the target column
• PCMCI:
  • Test choice CMI / ParCorr
  • Max lag 0,...,5
• TEFS:
  • Direction backward / forward
  • Use contemporary yes / no
  • Max lag features nolag,1,...,5
  • Max lag target 1,...,5

Outputs (causal):

• pickle dictionary with baseline
• PCMCI:
  • PDF table of selected features with on top the three R2 scores
  • pickle dictionary
• TEFS:
  • PDF table of selected features with on top the three R2 scores
  • pickle dictionary

Outputs (wrapper):

• Wrapper plot of the score evolution in all the configs
• pickle dictionary

More notes

• Assume data is already preprocessed
• Ask for train and test set separately

TODO List:

• Propagate user inputs "target column name"
• Raise an error when contemporary is false and lag is set to "no lag" in TEFS
• Test with wps_causal.py and make sure everything runs smoothly