| CPC G06F 16/285 (2019.01) [G06F 9/542 (2013.01); G06N 20/00 (2019.01)] | 20 Claims |

|
1. A method comprising:
receiving a data indication of a real-time event;
determining that the real-time event corresponds to an attribute of a user profile that is part of a plurality of features corresponding to an experience journey of a user, the experience journey comprising a set of journey states reflecting stages of interaction along a timeline;
determining, with a journey state machine-learning model and based on a feature of the real-time event, a journey state indicating a current stage of interaction of the user with respect to the experience journey, wherein the journey state machine-learning model comprises a machine-learning model iteratively trained by:
predicting journey states for training features; and
modifying the journey state machine-learning model based on comparing the journey states with ground truth journey states utilizing a loss function;
generating a future action score reflecting a likelihood that a user will discontinue use of a product or service based on the real-time event, the attribute of the user profile, and the journey state; and
determining a system action corresponding to the user profile based on the future action score.
|