search

mtreinish / ciml

a machine learning pipeline for analyzing CI results.
Apache License 2.0
5 stars 4 forks source link

Use a third dataset for predictions #21

Closed afrittoli closed 4 years ago

afrittoli commented 4 years ago

CIML supports splitting the dataset in three parts. We never really used the "dev" since there was no need for a smaller dataset for faster iterations.

We want to get predictions from the estimator so that we may use external tools to build the confusion matrix and draw further metrics. We can use the "dev" set for that.