MACHINE LEARNING MODEL AND ENCODER TO PREDICT ONLINE USER JOURNEYS
The subject technology identifies a series of journey event types in an online user journey, the event types including an impression event, an email event, a click event, and a website visit, and assigns an encoder to each event type. Using an assigned encoder, the technology encodes each event type to generate an encoded vector for each event type. The encoded vector is representative of at least a portion of the online user journey relating to that event type. The technology generates an encoded vector for each event type to create a set of encoded vectors, the set of encoded vectors including one or more of an impression event encoded vector, an email event encoded vector, a click event encoded vector, and a website visit encoded vector. The technology aggregates the set of encoded vectors to generate an output of the online user journey encoder, the output including a composite encoded user journey vector for modeling, transmits the output of the online user journey encoder to a user journey training model for training of the model and, using a trained model, generates an occurrence probability for at least one further event in the online user journey.
1 . A system for predicting online user activity comprising:
one or more processors; and
a memory storing instructions which, when executed by at least one processor in the one or more processors, cause the at least one processor to:
identify an online user journey including multiple event types;
encode each instance of a particular event type in the online user journey to generate an encoded vector for the particular event type;
aggregate the encoded vectors for each event type into a set of encoded vectors for modeling;
transmit the set of encoded vectors to a user journey training model for training of a user journey model;
append a theoretical further event to the online user journey to generate a test online user journey, the theoretical further event selected based on occurrence probabilities for multiple further events generated by the trained user journey model;
update the set of encoded vectors by adding a test encoded vector encoding the theoretical further event; and
evaluate the updated set of encoded vectors using the trained user journey model to generate an occurrence probability for at least one further event in the test online user journey.
2 . The system of claim 1 , wherein the user journey model includes a long short-term memory (LSTM).
3 . The system of claim 1 , wherein the at least one further event includes a conversion event.
4 . The system of claim 3 , wherein the processor is further configured to determine how the occurrence probability of the conversion event is affected by differing content items associated with one or more of the multiple journey event types.
5 . The system of claim 3 , wherein the processor is further configured to optimize an occurrence probability of the conversion event using differing content items associated with one or more of the multiple journey event types.
6 . The system of claim 1 , wherein the processor is further configured to sequence the encoded vectors in the updated set of encoded vectors in at least one of a time-based order and event type order; and
aggregate the sequenced encoded vectors to generate an output of an online user journey encoder, the output including a composite encoded user journey vector.
7 . The system of claim 6 , wherein the processor is further configured to augment the composite encoded user journey vector based on one or more of a media type, a timing, a media channels, an image, and a text or email content.
8 . The system of claim 1 , wherein the processor is further configured to map each instance of one or more of the multiple event types to a unique user id.
9 . A method of predicting online user activity, the method comprising:
identifying an online user journey including multiple event types;
encoding each instance of a particular event type in the online user journey to generate an encoded vector for the particular event type;
aggregating the encoded vectors for each event type into a set of encoded vectors for modeling;
transmitting the set of encoded vectors to a user journey training model for training of a user journey model;
appending a theoretical further event to the online user journey to generate a test online user journey, the theoretical further event selected based on occurrence probabilities for multiple further events generated by the trained user journey model;
updating the set of encoded vectors by adding a test encoded vector encoding the theoretical further event; and
evaluating the updated set of encoded vectors using the trained user journey model to generate an occurrence probability for at least one further event in the test online user journey.
10 . The method of claim 9 , wherein the user journey model includes a long short-term memory (LSTM).
11 . The method of claim 9 , wherein the at least one further event includes a conversion event.
12 . The method of claim 11 , further comprising determining how the occurrence probability of the conversion event is affected by differing content items associated with one or more of the multiple journey event types.
13 . The method of claim 11 , further comprising optimizing an occurrence probability of the conversion event using differing content items associated with one or more of the multiple journey event types.
14 . The method of claim 9 , further comprising sequencing the encoded vectors in the updated set of encoded vectors in at least one of a time-based order or event type order; and
aggregating the sequenced encoded vectors to generate an output of an online user journey encoder, the output including a composite encoded user journey vector.
15 . The method of claim 14 , further comprising augmenting the composite encoded user journey vector based on one or more of a media type, a timing, a media channels, an image, and a text or email content.
16 . The method of claim 9 , further comprising mapping each instance of one or more of the multiple event types to a unique user id.
17 . A non-transitory machine-readable medium comprising instructions which, when read by a machine, cause the machine to perform operations in a method of predicting online user activity, the operations comprising:
identifying an online user journey including multiple event types;
encoding each instance of a particular event type in the online user journey to generate an encoded vector for the particular event type;
aggregating the encoded vectors for each event type into a set of encoded vectors for modeling;
transmitting the set of encoded vectors to a user journey training model for training of a user journey model;
appending a theoretical further event to the online user journey to generate a test online user journey, the theoretical further event selected based on occurrence probabilities for multiple further events generated by the trained user journey model;
updating the set of encoded vectors by adding a test encoded vector encoding the theoretical further event; and
evaluating the updated set of encoded vectors using the trained user journey model to generate an occurrence probability for at least one further event in the test online user journey.
18 . The medium of claim 17 , wherein the at least one further event includes a conversion event.
19 . The medium of claim 18 , wherein the operations further comprise determining how the occurrence probability of the conversion event is affected by differing content items associated with one or more of the multiple journey event types.
20 . The medium of claim 18 , wherein the operations further comprise optimizing an occurrence probability of the conversion event using differing content items associated with one or more of the multiple journey event types.