[ML] Meta - Regression UI

[alvarezmelissa87]

• Create - general
  • Increase prominence of the link to the Advanced (JSON) editor (must have)
  • finalize Regression job type help text (must have)
  • replace modal with flyout (must have)
    • validate job type must be selected
    • disable advanced editor if job type not selected
  • Type dropdown for outlier/regression type (must have)
    • no default value
  • User can customize model memory limit in basic form (should have)
    • [ML] DF Analytics - auto-populate model_memory_limit #50714
  • Model memory limit should be auto-populated using the estimate_model_memory_limit endpoint *
    • [ML] DF Analytics - auto-populate model_memory_limit #50714
  • User can add a description for the job in the form and advanced editor (must have)
    • [ML] DF Analytics: add description field to job creation and display in jobs list #52217
  • [ ] disable the Create button in the advanced editor if there is a syntax error in the JSON - (removed as code editor already provides validation)
  • Display the job description in the expanded rows of the jobs lists (must have)
    • [ML] DF Analytics: add description field to job creation and display in jobs list #52217
  • Display the job description in the jobs lists table (could have, if sufficient space to retain existing columns)
    • [ML] DF Analytics: add description field to job creation and display in jobs list #52217
  • User can select to included/excluded fields in basic form - use explain API (should have)
    • [ML] DF Analytics job creation: Add 'excludes' input field to form #53856
  • don't allow Delete action if job is not stopped (must have)
• Regression 
  • Create
    • User can create regression job using basic form with the fields(must have):
      • User can enter job ID (must have)
      • User can enter source index (must have)
      • User can enter destination index (must have)
      • User can enter dependent_variable (must have)
        • Fetch options using field_caps api - grab all numerical fields as options
        • Ensure error message shows up correctly
      • User can enter training percent (must have)
    • [ ] User can enter prediction_field_name (could have)
  • Evaluate
    • Once the regression job has completed, the results of the job should be evaluated to provide the r_squared and mean_squared_error metrics (must have)
    • Once the regression job has completed, perform an evaluation of the job when viewing the stats on the job (could have)
    • The evaluation step should be run automatically (should have)
    • The user can view the r_squared and mean_squared_error metrics (must have)
    • Evaluate the training set to provide the ‘training error’ (isTraining: true) (must have)
    • Evaluate the testing set to provide the ‘generalization error’ (isTraining: false) (must have)
      - [ ] Add text to R squared tooltip to explain range e.g. 'Range 0-1, with values closer to 1 having best correlation.'
    • Add link to docs in top right corner of panel
      • [ML] DF Analytics Results: adds link to docs #54189
  • Explore
    • The user can view the r_squared and mean_squared_error metrics (must have)
    • [ ] The r_squared and mean_squared_error metrics should link to documentation with further information (a tooltip with a brief explanation as well) - (replaced by links to docs in top right of evaluate panel)
    • Display the r_squared and mean_squared_error metrics for the training set and testing set (must have)
    • Re-run the evaluation and display the r_squared and mean_squared_error metrics for subset of data queried by user using the table query bar *
      • [ML] DF Analytics Regression results: rerun evaluate endpoint for search bar queries #50991
    • Take into account customizable 'predicted_field' when fetching eval data
      • [ML] DF Regression: fix custom results_field and prediction_field_name not considered in eval config #48599
    • The user can explore the regression results in the ML UI (should have)
    • The user can view the regression results in a table (should have)
      • [ML] DataFrame Analytics: Regression results view #49667
      • Highlight the dependent field - the source value and the predicted value (should have)
      • Display whether the document is training data (should have)
      • The user can filter for training or test data (should have)
      • The user can sort the values in the table by column (should have)
      • Add a query bar (could have)
      • Display a label showing number of docs obtained (default 1000)

[elasticmachine]

Pinging @elastic/ml-ui (:ml)

[alvarezmelissa87]

For the User can select to included/excluded fields in basic form (should have) checklist item under Create - just want to confirm that we want to show included/excluded fields both in the basic form and in the advanced editor. Thanks! 😄
cc @sophiec20, @peteharverson

[alvarezmelissa87]

@peteharverson, @sophiec20 I've added a couple of items (marked by *) to this list - would you be up for confirming those? Thank you!

[sophiec20]

Yes - they are valid.

The r_squared and mean_squared_error metrics should link to documentation with further information (perhaps a tooltip with a brief explanation as well?) *

The brief description in a tooltip is useful. Can we remove "perhaps".

Re-run the evaluation and display the r_squared and mean_squared_error metrics for subset of data queried by user using the table query bar *

Please chat through this with the data team in terms of how they would use it. How does it look if there is no generalization error (i.e. if they select "training"). We should probably show the count of docs evaluated for test/train. Do we always want to apply the query or only sometimes? Probably always.