Hybrid process

The challenge

Content-based recommendation processes depend on metadata and collaborative recommendation depends on a user history. Furthermore, it is only collaborative recommendation processes that can issue recommendations even across various content. They, in turn, experience a problem if the user is not yet known to the system, because a user profile is required for generating recommendations (see "Recommendation algorithms").

The YOOCHOOSE solution

To enhance the quality of the overall system, it is therefore necessary to skillfully connect the different processes depending on the situation. The graphic illustrates the problem in the different situations for which YOOCHOOSE has developed several solution methods.

One of them is the hybrid solution in which all processes used in the YOOCHOOSE system can be connected to each other. For example, this allows the mitigation of the cold start problem from the user's perspective. As long as no profile has been created for a new user, recommendations are generated based on content-based and stereotype processes. As soon as user profiling has concluded over the course of usage, the recommendation processes employed for this user will also change automatically. In doing so, the weighting between the processes and the point in time from which the user profile is to be used should be freely configurable and can therefore be optimally adjusted to the customer situation.