In Recommendation Models
and Definition
you can do the following:
To create a model proceed as follows:
Choose Create
in Recommendation Models
.
Enter a Name
, choose a Model Type
, and enter a Description
.
For more information, see Recommendation Model Types.
Choose Save
.
The new model now appears in the Recommendation Models
table in Recommendation Models
.
Note
You can also use an existing model as the basis to create a new model by choosing Save As
while viewing or editing a model in Definition
or Preview Recommendations
. The Reference
panel of models that you create in this manner contain the original model name and model type details. You can display the original model in a new tab by choosing the model name.
For more information, see Recommendation Models.
Having created the model, you can now proceed to edit it as follows:
Choose the name of the model in the Recommendation Models
table in Recommendation Models
.
Definition
opens in a separate browser tab.
Note
You can only edit models that have a Draft
activation status in the Key Information
panel.
Choose Edit
from the options available at the bottom of the page.
A dialog box appears if any changes, either by you (without saving) or another user, have been made to the model. The dialog box prompts you to acknowledge that you are overwriting any changes that are currently being made to the model. If you change and save the model; the Administrative Data
panel reflects that you are the last person to change the model and when the changes were made.
Choose Add Step
to add the number of steps your model requires and configure the step settings.
Models consist of one or more steps that contain one or more algorithms. In models that contain more than one step, the steps are processed from left to right using the results returned by the previous step. The following step types are available:
Recommend
The recommend step type uses algorithms to produce a result set.
Re-Rank
The re-rank step type can use algorithms or rules to change the ranking (score) of a result set.
Note
The use of rules is restricted to Manage Recommendations
. For more information, see Rules.
Filter
The filter step type can use algorithms or rules to remove recommendations from a result set.
Note
The use of rules in filter step types is restricted to Manage Recommendations
. If a filter step contains several tasks, consider whether the Result Merge Type
in the step settings should be Intersection
.
Convert
The convert step type converts data supplied by a data source into something that is supplied to a filter or re-rank step.
For example, a convert step within a model can examine the content of a consumer’s cart and determine which product category the items in it belong to. Then, in an additional step, re-rank the recommendations according to the product categories of the items contained in the cart.
Note
The convert step can only be added to models and edited by business analysts using Recommendation Models
.
All of the algorithms contained in a step are executed in parallel. The combined results of the algorithms are merged according to one of the following Result Merge Type
step settings:
Weighted
The weighted merge type renders results in descending order according to the final score assigned the recommendations. The final score of each recommendation is influenced by the weight assigned to the algorithm it is rendered from.
Intersection
The intersection merge type only renders recommendations results that are common among all of the algorithms contained within the step.
Note
The intersection merge type does not take algorithm weight into consideration when rendering recommendations.
The step settings also allow you to configure a maximum number of results that a step can yield.
Choose Add Algorithm
and choose an Algorithm Type
to display a list of related algorithms.
The algorithms are associated to an Algorithm Type ID
and Type of Step
in Customizing for SAP Hybris Marketing
under .
Choose an algorithm and configure the settings.
The tabs that are displayed in the algorithm settings are rendered according to the algorithm parameters, data sources, and associated data source pre-filters that are defined in Customizing under Data Source Definition
, Algorithm Definition
, and Configure Algorithms
. For more information, see Algorithms.
Example
You can define the Generation Refresh Rate (in hours)
on the Parameters
tab of relevant algorithms to periodically automate the generation of the algorithms within a Pending
or Active
model.
Enter an Active On
date to define when the model becomes active.
You must provide an Active On
date before you can activate a model.
Choose Save
.
Once you save and generate a model, you can preview the recommendations the model returns. For more information, see Previewing Recommendations Returned by a Model.
In Definition
and Preview Recommendations
you can manage models by doing the following:
Managing the model activation status
You can manage model activation by choosing Activate
, Cancel Activation
, or Deactivate
.
For more information, see Understanding Model Statuses.
Deleting a model
You can only delete models that have a Draft
activation status in the Key Information
panel.