US 11,734,066 B2
Resource scheduling using machine learning
Jinchao Li, Redmond, WA (US); Yu Wang, Redmond, WA (US); Karan Srivastava, Seattle, WA (US); Jianfeng Gao, Woodinville, WA (US); Prabhdeep Singh, Newcastle, WA (US); Haiyuan Cao, Redmond, WA (US); Xinying Song, Bellevue, WA (US); Hui Su, Bellevue, WA (US); and Jaideep Sarkar, Redmond, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jan. 8, 2020, as Appl. No. 16/737,474.
Application 16/737,474 is a continuation of application No. 15/943,206, filed on Apr. 2, 2018, granted, now 10,579,423.
Prior Publication US 2020/0142737 A1, May 7, 2020
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 9/46 (2006.01); G06F 9/48 (2006.01); G06F 9/50 (2006.01); G06N 20/00 (2019.01); G06F 18/21 (2023.01)
CPC G06F 9/4887 (2013.01) [G06F 9/4881 (2013.01); G06F 9/5005 (2013.01); G06F 18/21 (2023.01); G06N 20/00 (2019.01)] 15 Claims
OG exemplary drawing
 
12. A method of scheduling a resource to perform a task by a scheduler system, the method comprising:
determine feature values of features of a task based on a predefined mapping of the features of the task to the feature values in a first memory device, the features of the task including a time-dependent feature and a time-independent feature, the time-dependent feature indicating a time the task is more likely to be successfully completed by a resource;
determining, by a trained and deployed gradient boost tree model and using a first current time and the determined feature values as input to the trained and deployed gradient boost tree model, a likelihood the resource will successfully complete the task within a time window;
determining a specified amount of time has elapsed since the determined likelihood was determined, and re-determining the likelihood the resource will successfully complete the task, the likelihood re-determined by the gradient boost tree model based on a second current time and the determined feature values;
receiving an indication, at a time before the specified amount of time has elapsed since the likelihood was re-determined at the second current time, that a new feature for the task is available;
re-determining the likelihood the resource will successfully complete the task, the likelihood re-determined by the gradient boost tree model based on a third current time, the feature values, and a feature value for the new feature;
scheduling the task to be performed by the resource based on the re-determined likelihood at the third current time resulting in a schedule; and
performing the task by the resource based on the schedule.