Anonymous cross-device, cross-channel, and cross-venue user identification using adaptive deep learning
Embodiments of the present invention provide systems, methods, and computer storage media for digital user identification across different devices, channels, and venues. Generally, digital interactions of a user can reveal a pattern of digital behavior that can be detected and assigned to the user, and a classifier can be learned to identify the user. Various types of digital interaction data may be utilized to identify a user, including device data, geolocation data associated with a user device, clickstream data or other attributes of web traffic, and the like. Anonymity can be provided by only utilizing behavioral-based user data. Digital interaction data can be encoded and fed into a multi-class classifier (e.g., deep neural network, support vector machine, random forest, k-nearest neighbors, etc.), with each user corresponding to a different class. New users can be detected and used to automatically grow a deep neural network to identify additional classes for the new users.
1. One or more computer storage media storing computer-useable instructions that, when used by one or more computing devices, cause the one or more computing devices to perform operations comprising:
accessing a digital interaction encoding that encodes digital interaction data for a digital interaction of a user into an encoding vector;
generating a predicted identification of the user by performing a multi-class classification of the digital interaction encoding with a deep neural network, a support vector machine, a random forest classifier, or a k-nearest neighbor classifier; and
providing the predicted identification.
2. The computer storage media of claim 1 , wherein the digital interaction data comprises only behavior-based data and the predicted identification of the user is anonymous.
3. The computer storage media of claim 1 , wherein each class of the multi-class classification corresponds to a different user of a plurality of known users, wherein generating the predicted identification comprises:
identifying, with the deep neural network, the user as a new user based on a determination that the multi-class classification fails to satisfy a similarity threshold with respect to the known users; and
automatically growing the deep neural network to accommodate a new class corresponding to the new user.
4. The computer storage media of claim 3 , wherein automatically growing the deep neural network comprises adding a neuron to an output layer of the deep neural network.
5. The computer storage media of claim 3 , the operations further comprising automatically retraining, based on a number of detected new classes, a multi-class classifier performing the multi-class classification.
6. The computer storage media of claim 1 , wherein the digital interaction data comprises finite values that replace missing values.
7. The computer storage media of claim 1 , wherein performing the multi-class classification of the digital interaction encoding is with the deep neural network.
8. A computerized method for digital user identification, the method comprising:
accessing a digital interaction encoding that encodes digital interaction data for a digital interaction of a new user into an encoding vector;
using the digital interaction encoding as an input into a deep neural network configured to perform a multi-class classification, wherein each class of the multi-class classification corresponds to a different user of a plurality of known users;
identifying a new class corresponding to the new user based on a determination that the multi-class classification fails to satisfy a similarity threshold with respect to the known users; and
automatically growing the deep neural network based on the new class.
9. The computerized method of claim 8 , wherein the digital interaction data comprises only behavior-based data.
10. The computerized method of claim 8 , wherein the determination that the multi-class classification fails to satisfy the similarity threshold comprises:
identifying a neuron from an output layer of the deep neural network with a strongest activation number; and
determining that the strongest activation number is below an activation threshold.
11. The computerized method of claim 8 , wherein automatically growing the deep neural network comprises:
adding a neuron to an output layer of the deep neural network; and
adding a new connection between the added neuron and each neuron in a previous layer of the deep neural network.
12. The computerized method of claim 11 , wherein the deep neural network is automatically retrained based on at least one of a number of new classes or a number of new digital interactions detected since a prior training of the deep neural network.
13. The computerized method of claim 11 , wherein the deep neural network is automatically retrained based on a threshold percentage of at least one of detected new users or detected new digital interactions to the known users.
14. The computerized method of claim 8 , wherein the digital interaction data comprises finite values that replace missing values.
15. A computer system comprising:
one or more hardware processors and memory configured to provide computer program instructions to the one or more hardware processors; and
a means for generating a predicted identification of a user by using the one or more hardware processors to perform a multi-class classification, with a deep neural network, a support vector machine, a random forest classifier, or a k-nearest neighbor classifier, based on a digital interaction encoding that encodes digital interaction data for a digital interaction of the user into an encoding vector.
16. The computer system of claim 15 , wherein the digital interaction data comprises only behavior-based data and the predicted identification of the user is anonymous.
17. The computer system of claim 15 , wherein each class of the multi-class classification corresponds to a different user of a plurality of known users, wherein the means for generating the predicted identification is configured to:
identify the user as a new user based on a determination that the multi-class classification fails to satisfy a similarity threshold with respect to the known users; and
automatically grow to accommodate a new class corresponding to the new user.
18. The computer system of claim 17 , wherein the means for generating the predicted identification is configured to automatically grow by adding a neuron.
19. The computer system of claim 17 , further comprising an adaptive modeling component configured to utilize the one or more hardware processors to automatically retrain the means for generating the predicted identification based on a number of detected new classes.
20. The computer system of claim 15 , wherein the digital interaction data comprises finite values that replace missing values.