New Functionality and Changes Introduced with Decision Optimizer 9.4
New Features and Changes
This release delivers enhancements in Tree Aware Optimization (TAO), reporting and general performance improvements.
Additionally, the Action Effect Modeling 2.4 releaseincludes the following.
New Features and Changes
New Features and Changes
This release delivers enhancements in Tree Aware Optimization (TAO), reporting and general performance improvements.
- Tree Aware Optimization (TAO) enhancements:
- Ability to use same decision key multiple times in tree template
- When updating tree template, allow for changing base tree file while keeping consistency constraint setup
- For a base tree template, allow branching base tree leaf nodes that are not at the deepest (i.e. allow expanding of non-terminal leaves into subtrees)
- Performance improvements of segment handling
- Report enhancements:
- Increased width and height of Efficient Frontier (EF) screen
- In Data Model view, enabled sorting on "Report Order" column
- In Swap Set Report, exported table now has rows and columns in same order to GUI
- General performance improvements (including segment handling across all scenarios and eligibility constraint segment validation)
- In JOBS tab, when clicking on executed scenario from JOBS tab, the scenario is opened automatically together with its project and shown in Summary Report view
- Validation Scenario now executes over the post-data-processing records, which is consistent with other scenario types
- In Swap Set report for an Exploratory Scenario, a separate table is now displayed for each respective sub-scenario
Additionally, the Action Effect Modeling 2.4 releaseincludes the following.
New Features and Changes
- In Data Load section, added message on the last processing option and execution time for Empty values treatment
- In Base Model, Holdout target is now capped according to the target bounds
- WoE formula selection is exported in Model Reports as "Standard Deviation", "Deviation from Mean", or "Binary"
- Added message alerting user if data load has failed
- Added option to replace empty values with defaults, and ignore records with any remaining empty values
- Summary statistics tables are computed separately for Development and Holdout samples, both for initial and processed data
- In run log for Read Data, we report the size and the (up to 20) indices of the empty records found
- In run log for Apply Variable Use and Process Empty, we also report the size and the (up to 20) indices of the ignored records
- Special values are counted in initial summary statistics and removed from processed statistics
- For each binning table, it is now possible to Sort bins correctly by the breakpoint
- Fixed computation of Action Effect training error metric: removed unnecessary division by sample weight and applied square root to report RMSE. For consistency, Base Model error and New Dataset errors are also reported as RMSE.
Updated 1 year ago
Version 3.0