Generic workflow for classification of highly imbalanced datasets using deep learning

Patent metadata

Field	Value
Publication number	US-11416748-B2
Application number	US-201916718524-A
Country	US
Kind code	B2
Filing date	Dec 18, 2019
Priority date	Dec 18, 2019
Publication date	Aug 16, 2022
Grant date	Aug 16, 2022

Abstract

Methods, systems, and computer-readable storage media for providing a binary classifier include receiving a biased dataset, the biased data set including a plurality of records, each record being assigned to a class of a plurality of classes, one class including a majority class, performing data engineering on at least a portion of the biased dataset to provide a revised dataset, providing a trained deep autoencoder (DAE) by training a DAE using only records assigned to the majority class from the revised dataset, the trained DAE including a binary classifier that classifies records into one of the majority class and a minority class, validating the trained DAE using validation data that is based on at least a portion of the biased dataset, and providing the trained DAE for production use within a production system.

First claim

What is claimed is:

1. A computer-implemented method for providing a binary classifier, the method being executed by one or more processors and comprising:
receiving a biased dataset, the biased dataset comprising a plurality of records, each record being assigned to a class of a plurality of classes, one class comprising a majority class;
performing data engineering on at least a portion of the biased dataset to provide a revised dataset;
providing a trained deep autoencoder (DAE) by training a DAE using only records assigned to the majority class from the revised dataset, the trained DAE comprising the binary classifier that classifies records into one of the majority class and a minority class;
validating the trained DAE using validation data that is based on at least a portion of the biased dataset; and
providing the trained DAE for production use within a production system.

2. The method of claim 1 , wherein the DAE comprises a first hidden layer having a different number of neurons than an input layer, and a second hidden layer having a different number of neurons than the first hidden layer.

3. The method of claim 2 , wherein the DAE further comprises a third hidden layer having a lower number of neurons than the second hidden layer and having a greater number of neurons than an output layer.

4. The method of claim 1 , wherein the data engineering comprises one of reducing a dimensionality of records and expanding a dimensionality of records in the at least a portion of the biased dataset.

5. The method of claim 1 , wherein the data engineering comprises scaling feature values of records in the at least a portion of the biased dataset.

6. The method of claim 1 , wherein the production use of the trained DAE comprises:
providing a record as input to the trained DAE;
receiving at least one value as output from the trained DAE, the at least one value being generated based on processing of the record through the trained DAE and representing an error in recreation of the record by the trained DAE; and
assigning the record to one of the majority class and the minority class based on the at least one value.

7. The method of claim 1 , wherein the validation data comprises records of the majority class and records of the minority class.

8. A non-transitory computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for providing a binary classifier, the operations comprising:
receiving a biased dataset, the biased dataset comprising a plurality of records, each record being assigned to a class of a plurality of classes, one class comprising a majority class;
performing data engineering on at least a portion of the biased dataset to provide a revised dataset;
providing a trained deep autoencoder (DAE) by training a DAE using only records assigned to the majority class from the revised dataset, the trained DAE comprising the binary classifier that classifies records into one of the majority class and a minority class;
validating the trained DAE using validation data that is based on at least a portion of the biased dataset; and

providing the trained DAE for production use within a production system.

9. The computer-readable storage medium of claim 8 , wherein the DAE comprises a first hidden layer having a different number of neurons than an input layer, and a second hidden layer having a different number of neurons than the first hidden layer.

10. The computer-readable storage medium of claim 9 , wherein the DAE further comprises a third hidden layer having a lower number of neurons than the second hidden layer and having a greater number of neurons than an output layer.

11. The computer-readable storage medium of claim 8 , wherein the data engineering comprises one of reducing a dimensionality of records and expanding a dimensionality of records in the at least a portion of the biased dataset.

12. The computer-readable storage medium of claim 8 , wherein the data engineering comprises scaling feature values of records in the at least a portion of the biased dataset.

13. The computer-readable storage medium of claim 8 , wherein the production use of the trained DAE comprises:
providing a record as input to the trained DAE;
receiving at least one value as output from the trained DAE, the at least one value being generated based on processing of the record through the trained DAE and representing an error in recreation of the record by the trained DAE; and
assigning the record to one of the majority class and the minority class based on the at least one value.

14. The computer-readable storage medium of claim 8 , wherein the validation data comprises records of the majority class and records of the minority class.

15. A system, comprising:
a computing device; and
a computer-readable storage device coupled to the computing device and having instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations for providing a binary classifier, the operations comprising:
receiving a biased dataset, the biased dataset comprising a plurality of records, each record being assigned to a class of a plurality of classes, one class comprising a majority class;
performing data engineering on at least a portion of the biased dataset to provide a revised dataset;
providing a trained deep autoencoder (DAE) by training a DAE using only records assigned to the majority class from the revised dataset, the trained DAE comprising the binary classifier that classifies records into one of the majority class and a minority class;
validating the trained DAE using validation data that is based on at least a portion of the biased dataset; and
providing the trained DAE for production use within a production system.

16. The system of claim 15 , wherein the DAE comprises a first hidden layer having a different number of neurons than an input layer, and a second hidden layer having a different number of neurons than the first hidden layer.

17. The system of claim 16 , wherein the DAE further comprises a third hidden layer having a lower number of neurons than the second hidden layer and having a greater number of neurons than an output layer.

18. The system of claim 15 , wherein the data engineering comprises one of reducing a dimensionality of records and expanding a dimensionality of records in the at least a portion of the biased dataset.

19. The system of claim 15 , wherein the data engineering comprises scaling feature values of records in the at least a portion of the biased dataset.

20. The system of claim 15 , wherein the production use of the trained DAE comprises:
providing a record as input to the trained DAE;
receiving at least one value as output from the trained DAE, the at least one value being generated based on processing of the record through the trained DAE and representing an error in recreation of the record by the trained DAE; and
assigning the record to one of the majority class and the minority class based on the at least one value.

Assignees

• Sap Se

Inventors

• Patil Ajinkya
• Farooqi Waqas Ahmad
• Fibich Jochim
• Schmidt Eckehard
• Jaehnisch Michael

Classifications

• G06N3/088 Primary

  Non-supervised learning, e.g. competitive learning · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0499

  Feedforward networks · CPC title