Real-time predictive recommendation system using per-set optimization
In general, embodiments of the present invention provide systems, methods and computer readable media configured to use a per-set level optimization of the rank order of promotions to be recommended to a consumer. In some embodiments, machine learning is used offline to generate a predictive diversity model that receives one or more similarity rank features associated with a promotion (e.g., category, price band) as input, and produces an output multiplier to be applied to the promotion's respective associated relevance score (e.g., a relevance score representing a prediction of the promotion's conversion rate without diversity features). At run time, per-set optimization of the ordering of a set of promotions is implemented by adjusting the respective associated relevance scores of the promotions using the diversity model and then re-ordering the set of promotions based on their respective adjusted relevance scores.
1. A computer-implemented method for per-set optimization of a rank order of a set of promotions to be recommended to a consumer, the method comprising:
calculating a diversity multiplier for a promotion in an ordered set of promotions based on a predictive diversity model and a diversity feature set associated with the ordered set of promotions, wherein the diversity multiplier represents an effect of the diversity feature set on likelihood of conversion of the promotion based on one or more attributes associated with the promotion;
calculating a relevance score for the promotion based on the diversity multiplier;
re-ordering the ordered set of promotions based on the relevance score for the promotion to generate a reordered set of promotions; and
transmitting a top N most highly ranked promotions from the reordered set of promotions based on respective relevance scores to a consumer device associated with the consumer.
2. The computer-implemented method of claim 1 , wherein the calculating the diversity multiplier is based on at least one per-set promotion feature associated with the diversity feature set.
3. The computer-implemented method of claim 1 , wherein the calculating the diversity multiplier is based on a dissimilarity between the promotion displayed on an electronic interface and one or more other promotions displayed on the electronic interface.
4. The computer-implemented method of claim 1 , wherein the calculating the diversity multiplier is based on a predicted effect of the diversity feature set on an assigned base relevance score of the promotion.
5. The computer-implemented method of claim 1 , wherein the calculating the diversity multiplier is based on at least one of a promotion category associated with the promotion, a price band associated with the promotion, and a location associated with the promotion.
6. The computer-implemented method of claim 1 , wherein the computer-implemented method further comprises:
generating the predictive diversity model based on a supervised learning process.
7. The computer-implemented method of claim 1 , wherein the computer-implemented method further comprises:
training the predictive diversity model based on historical data associated with a predictive relevance model that facilitates ordering of the ordered set of promotions.
8. A computer program product, stored on a computer readable medium, comprising instructions that when executed on one or more computers cause the one or more computers to perform operations implementing cross-sell promotions ranking, the operations comprising:
calculating a diversity multiplier for a promotion in an ordered set of promotions based on a predictive diversity model and a diversity feature set associated with the ordered set of promotions, wherein the diversity multiplier represents an effect of the diversity feature set on likelihood of conversion of the promotion based on one or more attributes associated with the promotion;
calculating a relevance score for the promotion based on the diversity multiplier;
re-ordering the ordered set of promotions based on the relevance score for the promotion to generate a reordered set of promotions; and
transmitting a top N most highly ranked promotions from the reordered set of promotions based on respective relevance scores to a consumer device associated with a consumer.
9. The computer program product of claim 8 , wherein the calculating the diversity multiplier is based on at least one per-set promotion feature associated with the diversity feature set.
10. The computer program product of claim 8 , wherein the calculating the diversity multiplier is based on a dissimilarity between the promotion displayed on an electronic interface and one or more other promotions displayed on the electronic interface.
11. The computer program product of claim 8 , wherein the calculating the diversity multiplier is based on a predicted effect of the diversity feature set on an assigned base relevance score of the promotion.
12. The computer program product of claim 8 , wherein the calculating the diversity multiplier is based on at least one of a promotion category associated with the promotion, a price band associated with the promotion, and a location associated with the promotion.
13. The computer program product of claim 8 , wherein the operations further comprise:
generating the predictive diversity model based on a supervised learning process.
14. The computer program product of claim 8 , wherein the operations further comprise:
training the predictive diversity model based on historical data associated with a predictive relevance model that facilitates ordering of the ordered set of promotions.
15. A system, comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations implementing per-set optimization of a rank order of a set of promotions to be recommended to a consumer, the operations comprising:
calculating a diversity multiplier for a promotion in an ordered set of promotions based on a predictive diversity model and a diversity feature set associated with the ordered set of promotions, wherein the diversity multiplier represents an effect of the diversity feature set on likelihood of conversion of the promotion based on one or more attributes associated with the promotion;
calculating a relevance score for the promotion based on the diversity multiplier;
re-ordering the ordered set of promotions based on the relevance score for the promotion to generate a reordered set of promotions; and
transmitting a top N most highly ranked promotions from the reordered set of promotions based on respective relevance scores to a consumer device associated with a consumer.
16. The system of claim 15 , wherein the calculating the diversity multiplier is based on a dissimilarity between the promotion displayed on an electronic interface and one or more other promotions displayed on the electronic interface.
17. The system of claim 15 , wherein the calculating the diversity multiplier is based on a predicted effect of the diversity feature set on an assigned base relevance score of the promotion.
18. The system of claim 15 , wherein the calculating the diversity multiplier is based on at least one of a promotion category associated with the promotion, a price band associated with the promotion, and a location associated with the promotion.