US 12,452,471 B2
Systems and methods for predicting disruptions in digital content system sessions
Prasanna Vijayanathan, Sunnyvale, CA (US); Alexander Christian Pavlakis, New York City, NY (US); and Andrew David Eichacker, Overland Park, KS (US)
Assigned to Netflix, Inc., Los Gatos, CA (US)
Filed by Netflix, Inc., Los Gatos, CA (US)
Filed on Nov. 30, 2023, as Appl. No. 18/524,197.
Prior Publication US 2025/0184551 A1, Jun. 5, 2025
Int. Cl. H04N 21/25 (2011.01); H04N 21/258 (2011.01)
CPC H04N 21/251 (2013.01) [H04N 21/25891 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
generating a device-specific feature, a geographic feature, and an application-level feature associated with a session between a client device and a digital content system;
applying a disruption prediction deep neural network to the device-specific feature, the geographic feature, and the application-level feature to generate a disruption prediction for the session;
determining contribution levels of the device-specific feature, the geographic feature, and the application-level feature to the disruption prediction by determining positive contribution levels and negative contribution levels for the device-specific feature, the geographic feature, and the application-level feature relative to the disruption prediction for the session; and
generating an attribution report based on the contribution levels of the device-specific feature, the geographic feature, and the application-level feature to the disruption prediction.