Evolutionary parameter optimization for selecting optimal personalized screen carousels
Systems and associated methods are described for providing content recommendations. The system selects a first plurality of subsets of content categories, each subset of content categories comprising a first number of content categories. The subsets are assigned reward scores based on content popularity and duplication. The subset are then iteratively modified to increase the rewards scores. If the reward scores are still low, the process is repeated by selecting a second plurality of subsets of content categories, each subset of content categories comprising a second number of content categories, different from first number.
1. A method for providing content recommendations, the method comprising:
generating a user interface that displays identifiers of content items of a plurality of content categories;
(a) selecting a first plurality of subsets of content categories, each subset of content categories comprising a first number of content categories;
(b) selecting recommended content items for each category of the first plurality of subsets of content categories;
(c) calculating a reward score for each respective subset of content categories of the first plurality of subsets of content categories based on a popularity score of content items and a number of duplicative content items in each respective subset of content categories;
(d) modifying the first plurality of subsets of content categories to maximize the reward scores;
(e) in response to determining that the rewards scores of the first plurality of subsets of content categories are below a threshold:
(f) selecting a second plurality of subsets of content categories, each subset of content categories comprising a second number of content categories, different from the first number;
(g) selecting a third plurality of subsets of content categories, each subset of content categories comprising a third number of content categories, different from the first and second numbers;
(h) repeating steps (b)-(d) for the second plurality of subsets of content categories and for the third plurality of subsets of content categories;
(i) repeating steps (f)-(h) until reward scores stop improving; and
(j) generating for display identifiers for recommended content items of a subset of content categories of one of the second plurality of subsets of content categories and the third plurality of subsets of content categories, wherein the subset has the highest reward score, and wherein the generating for display comprises rearranging display of identifiers on the user interface to display the identifiers for the recommended content items.
2. The method of claim 1 , wherein selecting recommended content items for each category of the first plurality of subsets of content categories comprises performing an API call to a recommendation engine.
3. The method of claim 1 , wherein the modifying the first plurality of subsets of content categories comprises:
removing and replacing a subset with the lowest score.
4. The method of claim 1 , wherein the modifying the first plurality of subsets of content categories comprises:
modifying a subset to remove a content category and add a different content category.
5. The method of claim 1 , wherein the modifying the first plurality of subsets of content categories comprises:
modifying a subset by merging content categories from two subsets.
6. The method of claim 1 , wherein calculating the reward score comprises:
calculating the popularity score of content items in a respective subset; and
increasing the reward score for the respective subset proportionally to the popularity score.
7. The method of claim 1 , wherein calculating a reward score comprises:
calculating an amount of duplicative content items in a respective subset; and
modifying the reward score of the respective subset in reverse proportion to the amount of duplicative content items in the respective subset.
8. The method of claim 1 , wherein calculating a reward score comprises:
identifying an amount of desired duplication in a subset;
calculating an amount of duplicative content items in a respective subset; and
modifying the reward score of the respective subset in proportion to how close the amount of duplicative content items in the respective subset is to the amount of desired duplication.
9. The method of claim 1 , wherein selecting recommended content items for each category of the first plurality of subsets of content categories comprises selecting recommended content for a future time.
10. A system for providing content recommendations, the system comprising:
control circuitry configured to:
generate a user interface that displays identifiers of content items of a plurality of content categories;
(a) select a first plurality of subsets of content categories, each subset of content categories comprising a first number of content categories;
(b) select recommended content items for each category of the first plurality of subsets of content categories;
(c) calculate a reward score for each respective subset of content categories of the first plurality of subsets of content categories based on a popularity score of content items and a number of duplicative content items in each respective subset of content categories;
(d) modify the first plurality of subsets of content categories to maximize the reward scores;
(e) in response to determining that the rewards scores of the first plurality of subsets of content categories are below a threshold:
(f) select a second plurality of subsets of content categories, each subset of content categories comprising a second number of content categories, different from the first number;
(g) select a third plurality of subsets of content categories, each subset of content categories comprising a third number of content categories, different from the first and second numbers;
(h) repeat steps (b)-(d) for the second plurality of subsets of content categories and for the third plurality of subsets of content categories;
(i) repeat steps (f)-(h) until reward scores stop improving; and
display circuitry configured to:
(j) generate for display identifiers for recommended content items of a subset of content categories of one of the second plurality of subsets of content categories and the third plurality of subsets of content categories, wherein the subset has the highest reward score, and wherein the display circuitry is configured to generate for display by rearranging display of identifiers on the user interface to display the identifiers for the recommended content items.
11. The system of claim 10 , wherein the control circuitry is configured to select recommended content items for each category of the first plurality of subsets of content categories by performing an API call to a recommendation engine.
12. The system of claim 10 , wherein the control circuitry is configured to modify the first plurality of subsets of content categories by:
removing and replacing a subset with the lowest score.
13. The system of claim 10 , wherein the control circuitry is configured to modify the first plurality of subsets of content categories by:
modifying a subset to remove a content category and add a different content category.
14. The system of claim 10 , wherein the control circuitry is configured to modify the first plurality of subsets of content categories by:
modifying a subset by merging content categories from two subsets.
15. The system of claim 10 , wherein the control circuitry is configured to calculate a reward score by:
calculating the popularity score of content items in a respective subset; and
increasing the reward score for the respective subset proportionally to the popularity score.
16. The system of claim 10 , wherein the control circuitry is configured to calculate a reward score by:
calculating an amount of duplicative content items in a respective subset; and
modifying the reward score of the respective subset in reverse proportion to the amount of duplicative content items in the respective subset.
17. The system of claim 10 , wherein the control circuitry is configured to calculate a reward score by:
identifying an amount of desired duplication in a subset;
calculating an amount of duplicative content items in a respective subset; and
modifying the reward score of the respective subset in proportion to how close the amount of duplicative content items in the respective subset is to the amount of desired duplication.
18. The system of claim 10 , wherein the control circuitry is configured to select recommended content items for each category of the first plurality of subsets of content categories by selecting recommended content for a future time.