US 11,657,320 B2
Using online engagement footprints for video engagement prediction
Seyedmohsen Jamali, Sunnyvale, CA (US); Samaneh Abbasi Moghaddam, Sunnyvale, CA (US); Ali Abbasi, Mountain View, CA (US); and Revant Kumar, Mountain View, CA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Feb. 26, 2019, as Appl. No. 16/286,465.
Prior Publication US 2020/0272937 A1, Aug. 27, 2020
Int. Cl. G06N 20/00 (2019.01); G06F 16/74 (2019.01); G06N 5/02 (2023.01); G06F 16/78 (2019.01); G06F 16/735 (2019.01); H04N 21/81 (2011.01); G06Q 30/0251 (2023.01); H04N 21/25 (2011.01)
CPC G06N 20/00 (2019.01) [G06F 16/735 (2019.01); G06F 16/74 (2019.01); G06F 16/7867 (2019.01); G06N 5/02 (2013.01); G06Q 30/0255 (2013.01); H04N 21/251 (2013.01); H04N 21/812 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving, from a plurality of client devices, a plurality of events, each event indicating a type of engagement of a video item from among a plurality of types of engagement, wherein a first event indicates a first type of engagement of the plurality of types of engagement and a second event indicates a second type of engagement of the plurality of types of engagement;
using one or more machine learning techniques to train a prediction model that is based on the plurality of events and a plurality of features that includes the plurality of types of engagement;
the plurality of features includes a first set of features for entities associated with a first impression count range and a second set of features for entities associated with a second impression count range;
determining a number of impressions of one or more video items by a particular entity associated with a content request;
in response to determining that the number of impressions is within the first impression count range, generating feature values for the set of features associated with the first impression count range;
in response to determining that the number of impressions is within the second impression count range, generating feature values for the set of features associated with the second impression count range;
based on an impression count range, identifying a plurality of entity feature values for the particular entity, wherein two or more entity feature values in the plurality of entity feature values correspond to two or more of the plurality of types of engagement;
generating a prediction based on the plurality of entity feature values and the prediction model; and
using the prediction to determine whether to select, from a plurality of content items, a particular content item that includes particular video.