Automatic rule generation for next-action recommendation engine
A system can recommend a next action for a user. A memory can store user data corresponding to the user and can include historic interaction points. A behavior pattern can be identified based on two or more interaction points stored in the user data. An intent of the user based on the behavior pattern can be identified. The intent can be based on a previous behavior pattern of another user. Several probabilities that the user will meet one or more objectives can be determined based on the intent. The probabilities can be scored using and used to assign a policy to the first user. A next action can be recommended based on the policy and executed with respect to the user. The outcome of the recommended next action can be stored to the user data.
1. A computer-implemented method using a variant hybrid contextual multi-armed bandit model, the method comprising:
receiving user data corresponding to a first user of a plurality of users, the user data stored in a memory storage device and including a plurality of historic interaction points;
identifying a first behavior pattern based on two or more interaction points of the plurality of historic interaction points;
identifying an intent of the first user based on the first behavior pattern, the intent based on a second behavior pattern of a second user of the plurality of users;
determining a plurality of probabilities that the first user will meet each of a plurality of objectives based on the intent;
scoring each of the plurality of objectives based on the plurality of probabilities;
assigning a policy from a plurality of policies to the first user based on the scoring, the policy based on a mapping between the user data and one or more actions of a plurality of actions associated with the policy;
outputting to the first user a recommended next action from the plurality of actions associated with the assigned policy;
receiving a new interaction point from the first user and responsive to the recommended next action;
updating, based on the received new interaction point from the first user, the user data, to at least one of
identify a new behavior pattern based on the new interaction point,
identify a new intent based on the new behavior pattern, and
assign a new policy from the plurality of policies based on the new intent; and
outputting to the first user, based on the updating of the user data, a new recommended next action from the plurality of actions associated with the new assigned policy.
2. The method of claim 1 , wherein the interaction points include interactions and non-interactions of the first user.
3. The method of claim 1 , wherein outputting the recommended next action further comprises prompting the first user to provide additional information.
4. The method of claim 1 , wherein outputting the recommended next action further comprises prompting the first user to complete a transaction.
5. The method of claim 1 , wherein the assigned policy corresponds to a first objective of the plurality of objectives having a greater probability that the first user will meet the first objective than a second objective of the plurality of objectives.
6. The method of claim 1 , wherein the assigned policy is a policy of a second type and is assigned based on determining that a policy of a first type could not be identified.
7. The method of claim 1 , wherein the outputting the recommended next action further comprises determining that the recommended next action is more suitable for the first user than another action of the plurality of actions.
8. The method of claim 1 , wherein a first objective of the plurality of objectives comprises two or more stages.
9. The method of claim 8 , further comprising rewarding the first user in response to advancing to a subsequent stage from a prior stage, the subsequent stage being progressively closer to fulfilling the objective than the prior stage.
10. A system comprising:
a memory storage device configured to store user data corresponding to a first user of a plurality of users and a plurality of historic interaction points;
one or more processors implementing a variant hybrid contextual multi-armed bandit model configured to:
identify, using a first model, a first behavior pattern based on two or more interaction points of the plurality of historic interaction points;
identify, using the first model, an intent of the first user based on the first behavior pattern, the intent based on a second behavior pattern of a second user of the plurality of users;
determine, using the first model, a plurality of probabilities that the first user will meet each of a plurality of objectives based on the intent;
score, using the first model, each of the plurality of objectives based on the plurality of probabilities;
assign, using the first model, a policy from a plurality of policies to the first user based on the scoring, the policy based on a mapping between the user data and one or more actions of a plurality of actions associated with the policy;
output to the first user, using a second model, a recommended next action from the plurality of actions associated with the assigned policy;
receive a new interaction point from the first user responsive to the recommended next action;
update, based on the received new interaction point from the first user, the user data, to at least one of
identify a new behavior pattern based on the new interaction point,
identify a new intent based on the new behavior pattern, and
assign a new policy from the plurality of policies based on the new intent; and
output to the first user, based on the updating of the user data, a new recommended next action from the plurality of actions associated with the new assigned policy.
11. The system of claim 10 , wherein the interaction points include interactions and non-interactions of the first user.
12. The system of claim 10 , wherein the outputting of the recommended next action prompts the user to provide additional information.
13. The system of claim 10 , wherein the outputting of the recommended next action prompts the user to complete a transaction.
14. The system of claim 10 , wherein the assigned policy corresponds to a first objective of the plurality of objectives having a greater probability that the user will meet the first objective than a second objective of the plurality of objectives.
15. The system of claim 10 , wherein the assigned policy is a policy of a second type and is assigned based on determining that a policy of a first type could not be identified.
16. The system of claim 10 , wherein the one or more processors are further configured to determine that the recommended next action is more suitable for the user than another action of the plurality of the actions.
17. The system of claim 10 , wherein a first objective of the plurality of objectives comprises two or more stages.
18. The system of claim 17 , wherein the one or more processors are further configured to: reward the user in response to advancing to a subsequent stage from a prior stage, the subsequent stage being progressively closer to fulfilling the objective than the prior stage.