US 12,430,569 B2
Systems and methods for providing media content recommendations
Kyle Miller, Durham, NC (US); Bryan S. Scappini, Cary, NC (US); and James W. Lent, Durham, NC (US)
Assigned to Adeia Guides Inc., San Jose, CA (US)
Filed by Adeia Guides Inc., San Jose, CA (US)
Filed on Feb. 23, 2022, as Appl. No. 17/678,713.
Application 17/678,713 is a continuation of application No. 16/370,101, filed on Mar. 29, 2019, granted, now 11,288,582.
Prior Publication US 2022/0180216 A1, Jun. 9, 2022
Int. Cl. G06F 16/9535 (2019.01); G06F 16/34 (2025.01); G06N 3/12 (2023.01); G06N 5/02 (2023.01); G06Q 30/0202 (2023.01); H04N 21/25 (2011.01)
CPC G06N 5/02 (2013.01) [G06N 3/12 (2013.01); G06Q 30/0202 (2013.01); H04N 21/251 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
(a) accessing a plurality of recommendation algorithms;
(b) generating a plurality of candidate weight combinations, wherein each candidate weight combination of the plurality of candidate weight combinations comprises, for each respective recommendation algorithm of the plurality of recommendation algorithms, a corresponding weight;
(c) using the plurality of candidate weight combinations to generate a first plurality of content item recommendations, wherein each respective content item recommendation of the first plurality of content item recommendations is generated using a respective candidate weight combination of the plurality of candidate weight combinations;
(d) receiving a first plurality of requests for content items of the first plurality of content item recommendations over a first period of time;
(e) determining, for each of the first plurality of requests received at (d), which candidate weight combination of the plurality of candidate weight combinations was used to generate the requested content item;
(f) generating, for each respective candidate weight combination of the plurality of candidate weight combinations, an evaluation metric based on a number of the first plurality of requests determined at (e) to correspond to a content item generated using the respective candidate weight combination of the plurality of candidate weight combinations;
(g) modifying the plurality of candidate weight combinations by replacing a candidate weight combination of the plurality of candidate weight combinations having a lowest evaluation metric with a new candidate weight combination, wherein the new candidate weight combination is generated based on combining two or more candidate weight combinations, of the plurality of candidate weight combinations, each having an evaluation metric exceeding a threshold;
(h) using the modified plurality of candidate weight combinations to generate a second plurality of content item recommendations, wherein each respective content item recommendation of the second plurality of content item recommendations is generated using a respective candidate weight combination of the modified plurality of candidate weight combinations;
(i) receiving a second plurality of requests for content items of the second plurality of content item recommendations over a second period of time;
(j) determining, for each of the second plurality of requests received at (i), which candidate weight combination of the modified plurality of candidate weight combinations was used to generate the requested content item;
(k) generating, for each respective candidate weight combination of the modified plurality of weight combinations, an updated evaluation metric based on a number of the second plurality of requests determined at (i) to correspond to a content item generated using the respective candidate weight combination of the modified plurality of candidate weight combinations;
(l) comparing the updated evaluation metrics of the modified plurality of weight combinations to historical evaluation metrics to determine whether the updated evaluation metrics of the modified plurality of weight combinations have improved by a threshold margin with respect to the historical evaluation metrics, wherein the historical evaluation metrics comprise at least the evaluation metrics generated at (f);
(m) based, at least in part, on determining that the updated evaluation metrics of the modified plurality of weight combinations have not improved by the threshold margin:
(m)(1) identifying a candidate weight combination of the modified plurality of candidate weight combinations having a highest evaluation metric;
(m)(2) using the identified candidate weight combination to generate a third plurality of content item recommendations; and
(m)(3) causing identifiers associated with the third plurality of content item recommendations to be generated for display.