Behavioral prediction for targeted end users

Patent metadata

Field	Value
Publication number	US-10210453-B2
Application number	US-201514828150-A
Country	US
Kind code	B2
Filing date	Aug 17, 2015
Priority date	Aug 17, 2015
Publication date	Feb 19, 2019
Grant date	Feb 19, 2019

Abstract

Behavioral prediction for targeted end users is described. In one or more example embodiments, a computer-readable storage medium has multiple instructions that cause one or more processors to perform multiple operations. Targeted selectstream data is obtained from one or more indications of data object requests corresponding to a targeted end user. A targeted directed graph is constructed based on the targeted selectstream data. A targeted graph feature vector is computed based on one or more invariant features associated with the targeted directed graph. A behavioral prediction is produced for the targeted end user by applying a prediction model to the targeted graph feature vector. In one or more example embodiments, the prediction model is generated based on multiple graph feature vectors respectively corresponding to multiple end users. In one or more example embodiments, a tailored opportunity is determined responsive to the behavioral prediction and issued to the targeted end user.

First claim

What is claimed is:

1. A device implemented at least partially in hardware to predict targeted end user behavior to facilitate opportunity tailoring, the device comprising:
a behavioral prediction engine implemented in the device at least partially in the hardware, the behavioral prediction engine including modules that operate to generate a prediction model from an input of selectstream data, the modules of the behavioral prediction engine comprising:
a selectstream data obtainment module configured to obtain from a storage device the selectstream data corresponding to multiple end users, the selectstream data indicative of data object interaction by a respective end user of the multiple end users in website environments;
a directed graph construction module configured to construct multiple directed graphs, each respective directed graph constructed based on the selectstream data corresponding to the respective end user of the multiple end users, each directed graph associated with one or more invariant features that translate to marketing analytics reflective of the data object interaction by the respective end user in the website environments;
a graph feature vector computation module configured to compute multiple graph feature vectors, each respective graph feature vector computed based on the one or more invariant features associated with a respective directed graph of the multiple directed graphs, at least two of the multiple directed graphs being isomorphic having a same set of the one or more invariant features;
a prediction model generation module configured to generate the prediction model based on the multiple graph feature vectors using a machine learning system, the prediction model configured to produce a behavioral prediction responsive to targeted selectstream data that corresponds to a targeted end user associated with a first isomorphic directed graph, the behavioral predication based in part on a second isomorphic directed graph associated with an end user of the multiple end users; and
a prediction model application module configured to apply the prediction model to a targeted graph feature vector that is derived from the targeted selectstream data to produce a behavioral prediction for the targeted end user, and issue a tailored opportunity to the targeted end user based on the behavioral prediction.

2. The device as described in claim 1 , wherein each directed graph of the multiple directed graphs is representative of a path of travel of a corresponding end user of the multiple end users as the corresponding end user traverses over multiple data objects.

3. The device as described in claim 1 , wherein the directed graph construction module is further configured to:
assign vertices of a given directed graph of the multiple directed graphs based on data objects traversed by a corresponding end user of the multiple end users; and
define respective directed edges of the given directed graph based on respective pairs of source data objects and destination data objects that are traversed by the corresponding end user, the pairs of source data objects and destination data objects identified in the selectstream data of the corresponding end user.

4. The device as described in claim 1 , wherein the one or more invariant features comprise at least one intuitive invariant feature that translates to a physical marketing concept and at least one abstract invariant feature that pertains to directed graph topology.

5. The device as described in claim 1 , wherein the graph feature vector computation module is further configured to:
compute multiple respective real values that correspond to multiple respective invariant features of the one or more invariant features for each directed graph of the multiple directed graphs.

6. The device as described in claim 1 , wherein the prediction model generation module is further configured to:
perform a clustering operation to separate the multiple graph feature vectors into multiple clusters based on one or more similarities between or among different ones of the multiple graph feature vectors.

7. The device as described in claim 1 , wherein the prediction model generation module is further configured to:
perform a training operation on a classifier using at least a portion of the multiple graph feature vectors as a training set of graph feature vectors with each graph feature vector of the at least a portion associated with a classification category.

8. A system implemented to predict targeted end user behavior to facilitate opportunity tailoring, the system comprising:
one or more computing devices that implement a behavioral prediction engine at least partially in hardware, the behavioral prediction engine of the one or more computing devices including modules that operate to generate a prediction model from an input of targeted selectstream data, the modules of the behavioral prediction engine configured to perform operations comprising:
obtaining the targeted selectstream data from one or more indications of data object requests corresponding to a targeted end user, the targeted selectstream data indicative of the data object requests by the targeted end user in website environments;
constructing a targeted directed graph based on the targeted selectstream data, the targeted directed graph associated with one or more invariant features that translate to marketing analytics reflective of the data object requests by the targeted end user in the website environments;
computing a targeted graph feature vector based on the one or more invariant features associated with the targeted directed graph, which is one of at least two isomorphic directed graphs that include the targeted graph feature vector;
producing a behavioral prediction for the targeted end user by applying the prediction model to the targeted graph feature vector of the at least two isomorphic directed graphs;
determining a tailored opportunity responsive to the behavioral prediction; and
issuing the tailored opportunity to the targeted end user.

9. The system as described in claim 8 , wherein the data object requests comprise requests for website pages that are each identified by a uniform resource locator (URL); and wherein the obtaining comprises:
receiving, from an end-user device of the targeted end user, the data object requests for the website pages that are each identified by a URL.

10. The system as described in claim 8 , wherein the behavioral prediction comprises at least one of: an indication that the targeted end user is likely to make a purchase, an indication that the targeted end user is likely to join a customer associative program, or an indication that the targeted end user is likely to abandon an electronic cart without completing a purchase.

11. The system as described in claim 8 , wherein the producing comprises:
identifying one or more graph feature vectors based on a similarity to the targeted graph feature vector, the one or more graph feature vectors corresponding to one or more end users that exhibited a particular behavioral attribute; and
assigning to the behavioral prediction the particular behavioral attribute.

12. The system as described in claim 11 , wherein a measure of the similarity comprises an L-p norm distance; and wherein the identifying comprises:
determining the L-p norm distance between the targeted graph feature vector and the one or more graph feature vectors.

13. The system as described in claim 8 , wherein the prediction model comprises a clustering-based prediction model; and wherein the producing comprises:
determining, using the clustering-based prediction model, a relations

Assignees

• Adobe Systems Inc
• Adobe Inc

Inventors

• Krishnamurthy Balaji
• Singla Tushar

Classifications

• G06N5/022 Primary

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N99/005

  Physics · mapped topic

• G06F17/30958

  Physics · mapped topic

• G06Q30/00

  Commerce · CPC title

• G06F17/30867

  Physics · mapped topic