日本語
You can use the Use Cases tab to keep track of your ML projects. This tab was designed to allow you to document and track your use cases over time. It allows you to add descriptions, assign deadlines, estimate ROI, add assets like datasets and deployments, and more.
Figure 1. Use Cases tab
You can click a use case to see its details and make changes if needed. To create a new use case, click the Add new use case link (at the top of the screen).
New use case
For the new use case, add a title and description of the problem you are trying to solve. You can add use case owners here as well and the number of events you want to predict.
Figure 2. New Use Case
Figure 3. Goal Dates
Figure 4. Value
Figure 5. Value Prioritization assessment grid
Figure 6. Value calculator and templates
Figure 7. Assets
Figure 8. Comments
Figure 9. Activity
Figure 10. Stage
Figure 11. Filter
Figure 12. Filtered for "Ideation" use cases
Figure 13. Filtered for "In Production" use cases
Figure 14. Value Tracker
Figure 2. New Use Case
Under Target dates you can add a date for each stage: Ideation, Queued, Data Prep and Modeling, Validating and Deploying, and In Production. You don’t have to add all of these at the beginning; you can add and edit dates as you progress through the project.
Figure 3. Goal Dates
Use case value
Another important component to track is the value of the use case. To do this, navigate to the Value tab and look for the grid on the left-hand side of the screen. The Y-axis indicates the “business impact” and the X-axis indicates the “feasibility.”
Figure 4. Value
Notice that along the bottom of the grid it says “Do not attempt”. This is because if there is no business impact, then it is likely not worth pursuing. Similarly, if you look at the leftmost band on the vertical axis, you can see that if there is value but the feasibility is “none,” then DataRobot recommends that you "Try to simplify" the problem.
Figure 5. Value Prioritization assessment grid
Ideally, you want to pursue use cases that are highly feasible and that have high business impact.
There is also a value estimator on this page. You can add a raw value or use the calculator. There is a calculator for both binary classification and regression problems.
Figure 6. Value calculator and templates
You can set the currency, indicate the level of improvement you want to make, the number of decisions made annually and how many are made incorrectly, as well as the value of each decision.
Use case assets, comments, and activity
When you have finished setting up the value estimate, click Save. Now you can click Assets and add assets to the use case. This includes datasets, DataRobot projects, deployment objects, and applications.
Figure 7. Assets
You can add comments under the Comments tab to track any developments with the project.
Figure 8. Comments
The Activity tab lets you track any changes to the use case. This serves as a record of everything that has occurred with the use case such as dates added, dates changed, assets added, comments and changes in project stage, etc.
Figure 9. Activity
You can set the current stage of the project in the status bar. (This bar is available on every tab of the use case.)
Figure 10. Stage
If you go back to the Use Cases tab you can see a Stage filter at the top. You can filter the use cases based on the stage you are interested in. For example, you can filter by use cases that are in the Ideation stage (Figure 11).
Figure 11. Filter
The contents of the page will change depending on which stage you selected to filter. For example in the Ideation stage you see value, business impact, and feasibility for those use cases.
Figure 12. Filtered for "Ideation" use cases
However, if you filter on use cases that are In Production, you will see the potential value, realized value, number of predictions, and deployment status for each of those use cases.
Figure 13. Filtered for "In Production" use cases
Value of use case in production
If you click on a model that is In Production, the Value tab shows a new pair of graphs. These tell you the realized value of the deployment over time and the number of predictions being made.
Figure 14. Value Tracker
More information
If you’re a licensed DataRobot customer, search the in-app Platform Documentation for Use Case Tracking.
For ideas on potential use cases for your own business problems, have a look at DataRobot Pathfinder and its library of over 150 use cases.
