Interpretation of Partial Dependence Plot (PDP)

ESSEC_Student
ESSEC_Student Registered Posts: 1

Hi!

I have created a Partial Dependence Plot and the graph has an initial sharp increase in the curve, before decreasing again. Is it possible to identify the rows in the dataset for which these initial high partial dependence value are predicted? Or, more generally speaking, is it possible to identify the rows in the dataset for which the partial dependence values are predicted?

This would be of big help. Thanks a lot in advance!

Tagged:

Answers

  • Tanguy
    Tanguy Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Neuron, Dataiku DSS Adv Designer, Registered, Dataiku DSS Developer, Neuron 2023 Posts: 113 Neuron

    Hi @ESSEC_Student
    ,

    I think what you are looking for are Individual Conditional Expectations (ICE for short but also called Ceteris Paribus Profiles) from which the Partial Dependence Plot (PDP) is actually derived.

    Capture d’écran 2023-06-08 214252.jpg

    source: https://ema.drwhy.ai/partialDependenceProfiles.html


    Unfortunately, ICE plots are not natively available in dataiku (it would be nice to optionnaly have them plotted over a PDP though!). You can however calculate them in the "individual explanations" page of a visual model.

    Hope this helps.

Setup Info
    Tags
      Help me…