Multi-tenant, metadata-driven recommendation system
A method for generating a model for recommendations from an item data set for a target data set includes embedding a set of targets from the target data set in a shared coordinate space using a first embedding function, embedding a first set of items from the item data set in the shared coordinate space using a second embedding function, selecting at least one target from the set of targets, and identifying a second set of items from the first set of items that are proximate to the at least one target as candidates from the recommendations.
1. A method for generating a model for recommendations from an item data set for a target data set, the method comprising:
embedding vectorized target data, representative of targets from the target data set, in a latent space using a first embedding function;
embedding a vectorized first set of item data, representative of a first set of items from the item data set, in the latent space using a second embedding function;
selecting at least one target data in the latent space;
identifying, based on proximity to the at least one selected target data in the latent space, a second set of items from the first set of items as candidates for recommendation;
scoring each item in the second set of items using a first scoring mechanism;
ranking each item according to a score for each item;
computing relevance metrics of each ranked item from the second set of items; and
comparing the relevance metrics with a baseline model.
2. The method of claim 1 , further comprising:
converting the target data set into a first set of numerical vectors;
converting the item data set into a second set of numerical vectors; and
performing negative sampling to identify negative interactions with the item data set.
3. The method of claim 1 , wherein the second set of items are identified as a pre-defined number of nearest neighbors to the at least one selected target in the latent space.
4. The method of claim 1 , further comprising:
adding items to the second set of items identified using a set of additional candidate models.
5. A non-transitory machine-readable storage medium that provides instructions that, if executed by a set of one or more processors, are configurable to cause the set of one or more processors to perform operations of a method for generating a model for recommendations from an item data set for a target data set, the operations comprising:
embedding vectorized target data, representative of targets from the target data set, in a latent space using a first embedding function;
embedding a vectorized first set of item data, representative of a first set of items from the item data set, in the latent space using a second embedding function;
selecting at least one target data in the latent space;
identifying, based on proximity to the at least one selected target data in the latent space, a second set of items from the first set of items as candidates for recommendation;
scoring each item in the second set of items using a first scoring mechanism;
ranking each item according to a score for each item;
computing relevance metrics of each ranked item from the second set of items; and
comparing the relevance metrics with a baseline model.
6. The non-transitory machine-readable storage medium of claim 5 , the operations further comprising:
converting the target data set into a first set of numerical vectors;
converting the item data set into a second set of numerical vectors; and
performing negative sampling to identify negative interactions with the item data set.
7. The non-transitory machine-readable storage medium of claim 5 , wherein the second set of items are identified as a pre-defined number of nearest neighbors to the at least one selected target in the latent space.
8. The non-transitory machine-readable storage medium of claim 5 , the operations further comprising:
adding items to the second set of items identified using a set of additional candidate models.
9. An apparatus comprising:
a set of one or more processors; and
a non-transitory machine-readable storage medium that provides instructions that, if executed by the set of one or more processors, are configurable to cause the apparatus to perform operations of a method for generating a model for recommendations from an item data set for a target data set, the operations comprising,
embedding vectorized target data, representative of targets from the target data set, in a latent space using a first embedding function,
embedding a vectorized first set of item data, representative of a first set of items from the item data set, in the latent space using a second embedding function,
selecting at least one target data in the latent space,
identifying, based on proximity to the at least one selected target data in the latent space, a second set of items from the first set of items as candidates for recommendation,
scoring each item in the second set of items using a first scoring mechanism,
ranking each item according to a score for each item,
computing relevance metrics of each ranked item from the second set of items, and
comparing the relevance metrics with a baseline model.
10. The apparatus of claim 9 , wherein the operations further comprise,
converting the target data set into a first set of numerical vectors;
converting the item data set into a second set of numerical vectors; and
performing negative sampling to identify negative interactions with the item data set.
11. The apparatus of claim 9 , wherein the second set of items are identified as a pre-defined number of nearest neighbors to the at least one selected target in the latent space.
12. The apparatus of claim 9 , wherein the operations further comprise,
adding items to the second set of items identified using a set of additional candidate models.