Recommendations Based on OU - Machine Learning

The below information outlines the algorithmic techniques that are used to calculate recommendations for system features which use machine learning to deliver recommendations to users:

Trending for Your Position

Summary Organizational Units (OUs) represent a group of many users that are included in a specific demographic, as defined by the organization. The Cornerstone MLP uses the data generated by these OUs to create recommendations specific to the group of users. For the Trending for Your Position carousel, recommendations are generated based on the Position OU.

For each Position OU, there is a group of employees that fall into one of the below categories:

  • Users that are currently in the position
  • Users that were previously in the position and moved to another position
  • Users that were previously in the position but left the company

From this group of users, the Cornerstone MLP analyzes the accumulated history of all the training ever taken by its members. It also analyzes the registration data accumulated over time. The end result of this analysis is an output of recommendations that are contextualized to the specific position.

Examples

Example 1:

The Cornerstone MLP finds that the majority of the users currently in a certain position have taken Training A. It also finds that many users that were previously in the position took Training B. The MLP would likely recommend these two training items for the users in the position today. It might, however, prioritize Training A, because it prioritizes training taken by users currently in the position.

Example 2:

There are certain instances in which a position either has insufficient members or not enough training registration data for those members (e.g. the position may be newly created or it may be a position that only contains a few users that do not interact with the portal). In this scenario, the Cornerstone MLP first considers position(s) higher in the OU hierarchy. It also considers the learner's division and location (and then neighboring positions, divisions, and locations), and then performs the same analysis to gather sufficient recommendations.

OU training data is prioritized accordingly by the MLP:

  • Users currently/previously in Position
  • Users currently/previously in Division
  • Users currently/previously in Location
  • Users currently/previously in neighboring Position X (closest Position in hierarchy)
  • Users currently/previously in neighboring Division X (closest Division in hierarchy)
  • Users currently/previously in neighboring Location X (closest Location in hierarchy)
  • Users currently/previously in neighboring Position Y (next closest Position in hierarchy)
  • Users currently/previously in neighboring Division Y (next closest Division in hierarchy)
  • Users currently/previously in neighboring Location Y (next closest Location in hierarchy)

The MLP analyzes training data according to this OU hierarchy until it finds enough training items (i.e. 500) to recommend for the user.

Position OU Not Available

In cases where the portal does not use the Position OU at all, the MLP analyzes more users in other OUs until it finds enough training (i.e. 500) to recommend for the user. For a portal that has never had a user in the Position OU, but utilizes the Division and Location OU, the following would be analyzed:

  • Users currently/previously in Position
  • Users currently/previously in Division
  • Users currently/previously in Location
  • Users currently/previously in neighboring Position X (closest Position in hierarchy)
  • Users currently/previously in neighboring Division X (closest Division in hierarchy)
  • Users currently/previously in neighboring Location X (closest Location in hierarchy)
  • Users currently/previously in neighboring Position Y (next closest Position in hierarchy)
  • Users currently/previously in neighboring Division Y (next closest Division in hierarchy)
  • Users currently/previously in neighboring Location Y (next closest Location in hierarchy)

Relevancy Scores

The MLP applies a Relevancy Score to each training item for a specific group of users. It first considers training registrations for the user’s position, then training registrations for the user’s division and location. This is because data gathered from the position OU is assigned a higher Relevancy Score for recommendations generated for the Trending for Your Position carousel.

Machine Learning Machine Learning Machine Learning Machine Learning Machine Learning Platform Machine Learning Platform MLP MLP MLP MLP