The Cornerstone Machine Learning Platform (MLP) applies some exclusion logic to the recommendations prior to sending them to Cornerstone Learning. This includes the following:
The Cornerstone MLP will only ever recommend the latest version of a training item. This means that if the MLP sees Version 2.0 as relevant to the user, but Version 5.0 is the latest version for the training (as of the start of the Cluster Compute), then the MLP assumes both versions are equivalent and will recommend Version 5.0 for the user.
The Cornerstone MLP will never recommend training that the user has taken before (i.e. training that is on the user's transcript).
When do these exclusions apply? The MLP exclusions are only applied based on the data received by the Cluster Compute. This means that if a new version (4.0) is created by an administrator between Cluster Computes, the MLP will not know that the newer version exists until the next Cluster Compute, so it will continue to recommend Version 3.0.
Cornerstone Learning applies exclusion logic to the recommendations after they are received from the Cornerstone MLP. This logic is a defense wall that ensures users are only recommended training that is relevant to them and that they have the ability to take.
The following training types are excluded from the recommendations returned by the MLP and do not display in the training carousels:
- Training flagged as Exclude From Recommendations via the Course Catalog
- Training not available to the user
- Inactive training
- Training that was already requested by or assigned to the user
- Training that is NOT in a Published status
- Training that is NOT the latest version
- Training that has recurrence enabled and is already on the user's transcript
- Child training that is not standalone
- Child training that is standalone training and on a user's transcript as Active
- Child Training that is standalone training and on a user's transcript as Not Yet Activated
These exclusions are applied immediately and do not need to process in the Cluster Compute. For example, if a training item is made inactive, it is immediately removed from recommendations.