Name Model ID Created
Block List

BlockList enables you to provide a list of items that you do not want in recommendation results.

Example: block_item_id_1,block_item_id_2,block_item_id_3

Delete
Upsale

Upsale enables you to always include items to return in recommendation results.

Example: upsale_item_id_1,upsale_item_id_2,upsale_item_id_3;2

Delete
White List

White List enables you to provide list of items where recommendation results must be chosen.

Example: whitelist_item_id_1,whitelist_item_id_2

Delete
Seed Block List

Per Seed Block List enables you to provide per item a list of items that cannot be returned as recommendation results.

Example: seed_id_1,seed_id_2;exclude_id_1,exclude_id_2

Delete
Feature Block List

Feature Block List enables you to provide a feature name and a list of feature values that recommendation results must not satisfy.

Example: feature_name;feature_value_1,feature_value2

Delete
Feature White List

Feature White List enables you to provide a feature name and a list of feature values that recommendation results must satisfy.

Example: feature_name;feature_value_1,feature_value2

Delete

Step 1: Add a catalog file (schema)


Catalog uploaded


×


Step 2: Add usage files (schema)


Uploaded files

×

Step 3: Create a build



ACTIVE BUILD BUILD ID TYPE STATUS DATE CREATED
() × ×
Select seed items below, or search by name.

Loading Catalog...

-- End of Catalog --

Page

Cart

Your shopping cart is empty!


Updating recommendations...

No Recommendations


Build Type
Description
Number of Model Iterations
Number of Model Dimensions
Item Cut Off Lower Bound
Item Cut Off Upper Bound
User Cut Off Lower Bound
User Cut Off Upper Bound
Use Features In Model
Modeling Feature List
()
Enable Feature Correlation
Reasoning Feature List
()
Allow Cold Item Placement
Enable U2I
FBT Support Threshold
Support Threshold
Co-occurrence Unit
FBT Minimal Score
FBT Max Item Set Size
FBT Similarity Function
Similarity Function
Enable Cold Item Placement:
Enable Cold-To-Cold Recommendations:
Allow Seed Items In Recommendations:
Enable Backfilling:
Enable User Affinity:
Enable Modeling Insights
Splitter Strategy
Test Percent
Random Seed
Split Date
Popular Item Benchmark Window (in days):
Number of Catalog Items
Number of Catalog Items in Usage
Number of Users
Number of Usage Records
Catalog Coverage
Number of Catalog Items in Build
Number of Users in Build
Number of Usage Records in Build
Catalog Coverage in Build

Diversity (User Recommendations)

Diversity measures the distribution of items recommended. Each percentile bucket is represented by a span (min/max values that range between 0 and 100). The items close to 0 are the least popular. For instance, if the percentage value for the 99-100 percentile bucket is 10.6, it means that 10.6 percent of the recommendations returned only the top 1% most popular items. The percentile bucket min value is inclusive, and the max value is exclusive except for 100.

Total Items Recommended in Test Set
Unique Items Recommended in Test Set
Total Items in Train Set

Precision at K (User Recommendations)

K represents the number of recommendations shown to the customer. So, if the Percentage under 5 is 4.94, the table would reads as follows: "if during the test period, only 5 user-based recommendations would have been shown to the customers, 4.94 of the users would have actually purchased at least one recommended item".

K
Precision (User Recommendations)
Precision (Most Popular Items)
Users in test
Users considered
Users not considered

Diversity (Item Recommendations)

Diversity measures the distribution of items recommended. Each percentile bucket is represented by a span (min/max values that range between 0 and 100). The items close to 0 are the least popular. For instance, if the percentage value for the 99-100 percentile bucket is 10.6, it means that 10.6 percent of the recommendations returned only the top 1% most popular items. The percentile bucket min value is inclusive, and the max value is exclusive except for 100.

Total Items Recommended in Test Set
Unique Items Recommended in Test Set
Total Items in Train Set

Precision at K (Item Recommendations)

K represents the number of recommendations shown to the customer. So, if the Percentage under 5 is 4.94, the table would reads as follows: "if during the test period, only 5 item-based recommendations would have been shown to the customers, 4.94 of the users would have actually purchased at least one recommended item".

K
Precision (Item Recommendations)
Precision (Most Popular Items)
Users in test
Users considered
Users not considered

To view Offline Metrics, create a build with modeling insights enabled.

Feature Name Rank