Performance indicators in Auto modelling

Danijela
Danijela New Altair Community Member
edited November 5 in Community Q&A
Would anyone be so kind to tell me why the percentage of prediction accuracy and the value of r2 are not appearing in my automodel results?
Is it possible to add the value of r2 in the prediction chart?

Thank you in advance!
Tagged:

Best Answer

  • jacobcybulski
    jacobcybulski New Altair Community Member
    edited December 2020 Answer ✓
    By default Auto Model will not use r2 performance measure. Note that r2 is just squared correlation r^2, which is equal to coefficient of determination R2 (I assume you are after this) only when you apply the model to training data, otherwise it is a generalisation of R2, as estimated for the population. As you already have correlation measure displayed, you will not get much more from r2 when displayed side by side for different models.

    However, if you insist on using other, not supported, types of measurements for your automatically generated estimation models (which is a bit scary), you can save each model once the Auto Model completes its run (see the big round button called "Open Process" for each of the results). Once you save the model, you can modify section (4) "Validate Model" group, then go to Performance for Hold-Out sets, and change the performance settings, and run each model separately.

Answers

  • jacobcybulski
    jacobcybulski New Altair Community Member
    edited December 2020 Answer ✓
    By default Auto Model will not use r2 performance measure. Note that r2 is just squared correlation r^2, which is equal to coefficient of determination R2 (I assume you are after this) only when you apply the model to training data, otherwise it is a generalisation of R2, as estimated for the population. As you already have correlation measure displayed, you will not get much more from r2 when displayed side by side for different models.

    However, if you insist on using other, not supported, types of measurements for your automatically generated estimation models (which is a bit scary), you can save each model once the Auto Model completes its run (see the big round button called "Open Process" for each of the results). Once you save the model, you can modify section (4) "Validate Model" group, then go to Performance for Hold-Out sets, and change the performance settings, and run each model separately.

Categories