User classification using a deep forest network
The present disclosure describes techniques for object classification using deep forest networks. One example method includes classifying a user object including features associated with the user based on a deep forest network including identifying one or more user static features, one or more user dynamic features, and one or more user association features from the features included in the user object; providing the user static features to first layers, the user dynamic features to second layers, and the user association features to third layers, the first, second, and third layers being different and each providing classification data to the next layer based at least in part on the input data and the provided user features.
1. A computer-implemented method for detecting anomalous user behaviors, the computer-implemented method comprising:
classifying a user object based on a deep forest network, the user object representing a user and including information associated with the user, the deep forest network including a plurality of layers each configured to receive input data from a previous layer and provide classification data to a next layer, wherein classifying the user object includes:
partitioning features into a plurality of feature groups comprising a user static feature group, a user dynamic feature group, and a user association feature group, wherein the features are extracted from the information associated with the user;
repeatedly providing the user static features to a first layer in every three layers, each first layer configured to provide the classification data to the next layer based at least in part on the input data and the user static features;
repeatedly providing the user dynamic features to a second layer in every three layers, each second layer configured to provide the classification data to the next layer based at least in part on the input data and the user dynamic features;
repeatedly providing the user association features to a third layer in every three layers, each third layer configured to provide the classification data to the next layer based at least in part on the input data and the user association features; and
obtaining final prediction results, the final prediction results including probability scores for each of a plurality of categories, each probability score representing a predicted likelihood that the user object belongs to the category.
2. The computer-implemented method of claim 1 , wherein classifying the user object based on the deep forest network includes:
providing features belonging to the user static feature group to one or more base classifiers, each base classifier configured to provide an initial classification of the user object as input to an initial layer from the plurality of layers.
3. The computer-implemented method of claim 1 , wherein obtaining final prediction results comprises:
averaging final classification results that are generated by a last layer of the plurality of layers.
4. The computer-implemented method of claim 1 , wherein the user static feature group includes at least one feature representing gender, race, or nationality.
5. The computer-implemented method of claim 1 , wherein the user dynamic feature group includes features representing a set of transactions performed by the user represented by the user object.
6. The computer-implemented method of claim 1 , wherein the user association feature group includes features representing a set of other users associated with the user represented by the user object.
7. The computer-implemented method of claim 1 , wherein classifying the user object based on the deep forest network includes classifying the user represented by the user object as normal or abnormal.
8. A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:
classifying a user object based on a deep forest network, the user object representing a user and including information associated with the user, the deep forest network including a plurality of layers each configured to receive input data from a previous layer and provide classification data to a next layer, wherein classifying the user object includes:
partitioning features into a plurality of feature groups comprising a user static feature group, a user dynamic feature group, and a user association feature group, wherein the features are extracted from the information associated with the user;
repeatedly providing the user static features to a first layer in every three layers, each first layer configured to provide the classification data to the next layer based at least in part on the input data and the user static features;
repeatedly providing the user dynamic features to a second layer in every three layers, each second layer configured to provide the classification data to the next layer based at least in part on the input data and the user dynamic features;
repeatedly providing the user association features to a third layer in every three layers, each third layer configured to provide the classification data to the next layer based at least in part on the input data and the user association features; and
obtaining final prediction results, the final prediction results including probability scores for each of a plurality of categories, each probability score representing a predicted likelihood that the user object belongs to the category.
9. The non-transitory, computer-readable medium of claim 8 , wherein classifying the user object based on the deep forest network includes:
providing features belonging to the user static feature group to one or more base classifiers, each base classifier configured to provide an initial classification of the user object as input to an initial layer from the plurality of layers.
10. The non-transitory, computer-readable medium of claim 8 , wherein obtaining final prediction results comprises:
averaging final classification results that are generated by a last layer of the plurality of layers.
11. The non-transitory, computer-readable medium of claim 8 , wherein the user static feature group includes at least one feature representing gender, race, or nationality.
12. The non-transitory, computer-readable medium of claim 8 , wherein the user dynamic feature group includes features representing a set of transactions performed by the user represented by the user object.
13. The non-transitory, computer-readable medium of claim 8 , wherein the user association feature group includes features representing a set of other users associated with the user represented by the user object.
14. The non-transitory, computer-readable medium of claim 8 , wherein classifying the user object based on the deep forest network includes classifying the user represented by the user object as normal or abnormal.
15. A computer-implemented system, comprising:
one or more computers; and
one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform one or more operations comprising:
classifying a user object based on a deep forest network, the user object representing a user and including information associated with the user, the deep forest network including a plurality of layers each configured to receive input data from a previous layer and provide classification data to a next layer, wherein classifying the user object includes:
partitioning features into a plurality of feature groups comprising a user static feature group, a user dynamic feature group, and a user association feature group, wherein the features are extracted from the information associated with the user;
repeatedly providing the user static features to a first layer in every three layers, each first layer configured to provide the classification data to the next layer based at least in part on the input data and the user static features;
repeatedly providing the user dynamic features to a second layer in every three layers, each second layer configured to provide the classification data to the next layer based at least in part on the input data and the user dynamic features;
repeatedly providing the user association features to a third layer in every three layers, each third layer configured to provide the classification data to the next layer based at least in part on the input data and the user association features; and
obtaining final prediction results, the final prediction results including probability scores for each of a plurality of categories, each probability score representing a predicted likelihood that the user object belongs to the category.
16. The system of claim 15 , wherein classifying the user object based on the deep forest network includes:
providing features belonging to the user static feature group to one or more base classifiers, each base classifier configured to provide an initial classification of the user object as input to an initial layer from the plurality of layers.
17. The system of claim 15 , wherein obtaining final prediction results comprises:
averaging final classification results that are generated by a last layer of the plurality of layers.
18. The system of claim 15 , wherein the user static feature group includes at least one feature representing gender, race, or nationality.
19. The system of claim 15 , wherein the user dynamic feature group includes features representing a set of transactions performed by the user represented by the user object.
20. The system of claim 15 , wherein the user association feature group includes features representing a set of other users associated with the user represented by the user object.