Systems and methods for evaluating models that generate recommendations
A device may receive content data, a first model, and a second model. The first model may be trained on different types of metadata than the second model. The content data may include a first identifier of a first content item and a first set of metadata associated with the first content item. The device may process the first set of metadata to generate first recommendations from the first model and second recommendations from the second model. The device may provide the first identifier and a combination of the first recommendations and the second recommendations to client devices. The device may receive, from the client devices, user-generated target recommendations based on the combination. The device may process the user-generated target recommendations, the first recommendations, and the second recommendations, to provide feedback to update the first model and the second model.
1. A method, comprising:
processing, by a device and within a testing mode, metadata to generate a plurality of recommendations from a plurality of models;
receiving, by the device, user-generated target recommendations based on the plurality of recommendations;
processing, by the device and within the testing mode, the user-generated target recommendations and the plurality of recommendations to determine a plurality of performance scores associated with the plurality of models and provide feedback for updating the plurality of models;
comparing, by the device and within the testing mode, a first one of the plurality of performance scores, associated with a first one of the plurality of models, to a second one of the plurality of performance scores associated with a second one of the plurality of models; and
causing, by the device, the plurality of models to be updated based on the comparing and the feedback.
2. The method of claim 1 , further comprising:
receiving a number of inputs;
calculating a plurality of conversion values for the plurality of models based on the number of inputs; and
updating the plurality of performance scores based on the plurality of conversion values.
3. The method of claim 1 , wherein processing the user-generated target recommendations and the plurality of recommendations comprises:
determining a plurality of distance metrics associated with the user-generated target recommendations and the plurality of recommendations;
calculating a plurality of weighted scores for the plurality of recommendations based on the plurality of distance metrics; and
calculating the plurality of performance scores based on the plurality of weighted scores.
4. The method of claim 1 , further comprising:
deleting a model of the plurality of models; or
generating a report relating to a performance score, of the plurality of performance scores, associated with the model.
5. The method of claim 1 , wherein each of the plurality of models are trained based on different types of metadata.
6. The method of claim 1 , wherein the user-generated target recommendations include at least one of the plurality of recommendations.
7. The method of claim 1 , further comprising:
reevaluating a model, of the plurality of models, after the model has been updated based on the feedback and the plurality of performance scores.
8. A device, comprising:
one or more processors configured to:
process, within a testing mode, metadata to generate a plurality of recommendations from a plurality of models;
receive user-generated target recommendations based on the plurality of recommendations;
process, within the testing mode, the user-generated target recommendations and the plurality of recommendations to determine a plurality of performance scores associated with the plurality of models;
compare, within the testing mode, a first one of the plurality of performance scores, associated with a first one of the plurality of models, to a second one of the plurality of performance scores associated with a second one of the plurality of models; and
cause an action to be performed based on the comparison.
9. The device of claim 8 , wherein the one or more processors are further configured to:
present a combination of recommendations, of the plurality of recommendations, in a particular order for display on a user interface.
10. The device of claim 8 , wherein each of the plurality of models is trained based on different types of metadata.
11. The device of claim 8 , wherein the one or more processors, to process the user-generated target recommendations and the plurality of recommendations, are configured to:
determine a plurality of distance metrics associated with the user-generated target recommendations and the plurality of recommendations;
calculate a plurality of weighted scores for the plurality of recommendations based on the plurality of distance metrics; and
calculate the plurality of performance scores based on the plurality of weighted scores.
12. The device of claim 8 , wherein the one or more processors are further configured to:
reevaluate a model, of the plurality of models, after the model has been updated based on feedback and the plurality of performance scores.
13. The device of claim 8 , wherein the one or more processors are further configured to:
delete a model of the plurality of models.
14. The device of claim 8 , wherein the one or more processors, to compare the first one of the plurality of performance scores to the second one of the plurality of performance scores, are configured to:
compare the first one of the plurality of performance scores to the second one of the plurality of performance scores to determine which model exceeds a user-generated target recommendation threshold.
15. A non-transitory computer-readable medium storing instructions, the instructions comprising:
one or more instructions that, when executed by one or more processors, cause the one or more processors to:
process, within a testing mode, metadata to generate a plurality of recommendations from a plurality of models;
receive user-generated target recommendations based on the plurality of recommendations;
process, within the testing mode, the user-generated target recommendations and the plurality of recommendations to determine a plurality of performance scores associated with the plurality of models and provide feedback for updating the plurality of models; and
cause an action to be performed based on the plurality of performance scores and the feedback for updating the plurality of models.
16. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:
compare a first one of the plurality of performance scores, associated with a first one of the plurality of models, to a second one of the plurality of performance scores associated with a second one of the plurality of models; and
wherein the one or more instructions, that cause the one or more processors to cause the action to be performed, cause the one or more processors to:
cause the action to be performed based on the comparison.
17. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:
reevaluate a model, of the plurality of models, after the model has been updated based on the feedback and the plurality of performance scores.
18. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:
determine a plurality of distance metrics associated with the user-generated target recommendations and the plurality of recommendations;
calculate a plurality of weighted scores for the plurality of recommendations based on the plurality of distance metrics; and
calculate the plurality of performance scores based on the plurality of weighted scores.
19. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:
generate a report based on plurality of performance scores; and
provide the report for display.
20. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:
calculate a plurality of conversion values for the plurality of models based on a number of inputs; and
update the plurality of performance scores based on the plurality of conversion values.