search

EpistasisLab / tpot

A Python Automated Machine Learning tool that optimizes machine learning pipelines using genetic programming.
http://epistasislab.github.io/tpot/
GNU Lesser General Public License v3.0
9.67k stars 1.56k forks source link

Accessing predictions of all pipelines ran. #829

Open aidiss opened 5 years ago

aidiss commented 5 years ago

Is it possible to see how each individual pipeline performed on the row level? I would like to access each individual prediction.

weixuanfu commented 5 years ago

evaluated_individuals_ attribute (see TPOT API) stores some statistics, like average CV score, of all evaluated pipelines, but the prediction target for X is not stored.

aidiss commented 5 years ago

With what classes could I mess to export the prediction targets? What module should I look into?

weixuanfu commented 5 years ago

There is no easy way in TPOT to export the prediction targets of all evaluated pipelines since TPOT won't generate them during optimization. The fitness score in TPOT is average CV score with spliting training set via (Stratified)KFold