Overview in Decision Optimization Experiment
This post highlights the main characteristics of the recently introduced Overview in Decision Optimization Experiment, which displays the main characteristics of all scenarios in a single table.
There are two alternatives to create Decision Optimization (DO) model formulations in the Watson Studio (WS) part of Cloud Pak for Data (CP4D). One way is to use Jupyter Notebooks and the other is to use DO Experiment (previously known as model builder).
In short, benefits of DO Experiments are to easily (without coding) use multiple different scenarios (with different model formulations and/or different data sets) and visualizations (charts and tables with multipage layouts). Those two functionalities are critical when developing DO model formulations. Also DO Experiment supports OPL and natural language Modeling Assistant, in addition to Python.
Why work with multiple scenarios?
So you have created a first formulation of a DO model which seems to solve and provide the right solution and you think you are done?
Is it working correctly with all types of input data or it breaks when some value is missing or is too high? You will need to test your formulation with different posible data sets and robustify your model for these particular cases.
Is it returning the right level of solution quality in the expected amount of time? You will need to tune your model formulations, and try some variations.
Is it providing the right optimal solutions and no constraints are missing? You will need to validate the model with a Business Analytst.
In all these cases, you will use multiple scenarios, with same or different model formulations and same or different data sets.
After you have created all these scenarios in a DO Experiment, you will be able to find out which one is best to deploy. For this you might need to see all the scenarios in one table. This is what the Overview is about.
You can open it by clicking on Overview in the top left.
The Overview will show all scenarios.
Overview features
In the top, you can see the number of scenarios according to their completion status.
A set of default columns are shown, but you can configure the ones of interest using the gear on the top right of the table.
You can also filter out columns using the Search table
text box and sort the rows according to a given column.
This way, you can easily find out the fastest model or the one with the smaller gap within a given time limit.
The Overview also allows to run actions on a set of scenarios. For now this is limited to deleting these scenarios.
The Overview is available in the public cloud and in the CP4D 3.5 release.
For more stories about AI and DO, follow me on Medium, Twitter or LinkedIn.