US 12,455,750 B2
Machine learning for predicting next best action
Yao Chen, San Jose, CA (US); Lingjie Weng, Sunnyvale, CA (US); Arvind Murali Mohan, Sunnyvale, CA (US); Hongbo Zhao, San Jose, CA (US); Lu Chen, Sunnyvale, CA (US); Dipen Thakkar, San Jose, CA (US); Xiaoxi Zhao, Milpitas, CA (US); Shifu Wang, San Jose, CA (US); Jim Chang, Cupertino, CA (US); Daniel D Thorndyke, San Jose, CA (US); and Smriti R. Ramakrishnan, Belmont, CA (US)
Assigned to MICROSOFT TECHNOLOGY LICENSING, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jun. 9, 2023, as Appl. No. 18/208,199.
Prior Publication US 2024/0411573 A1, Dec. 12, 2024
Int. Cl. G06F 9/451 (2018.01)
CPC G06F 9/453 (2018.02) 20 Claims
OG exemplary drawing
 
1. A system comprising:
at least one processor;
a non-transitory computer-readable medium having instructions stored thereon, which, when executed by the at least one processor, cause the system to perform operations comprising:
receiving an indication of an action performed in a graphical user interface at direction of a first user;
obtaining one or more signals regarding the first user and a request context of the action, the request context indicating a context of the graphical user interface in which the action was performed, the request context comprising at least one of intra-page context including a number of impressions taken by the first user on a current graphical user interface page, intra-session context including a number of actions taken and a type of action taken by the first user within a current session, or inter-session context including a number of actions of different types which the first user took over a predetermined number of prior sessions;
retrieving, via an Application Program Interface (API) layer request, client-side navigation commands comprising a set of user interface elements currently displayed to the user after performing the action;
obtaining, using the request context and the set of user interface elements currently displayed to the user, eligible actions comprising a list of next possible graphical user interface actions to be performed by the user;
fetching, via a feature retrieval component, feature data for the eligible actions;
training a deep machine learning model on actions comprising entity level impressions made to users and request contexts to output, for a given user and request context, next possible graphical user interface actions to present to the user;
feeding the one or more signals and the feature data for the eligible actions into the deep machine learning model, the deep machine learning model outputting one or more recommended next possible graphical user interface actions from the list of next possible graphical user interface actions based on inferred intent of the first user from the one or more signals, the deep machine learning model trained to output recommendations that optimize a plurality of different performance metrics by scoring only the eligible actions and ranking the eligible actions according to their respective scores; and
causing display of the one or more recommended next possible graphical user interface actions.