Back to results
US 12,406,261 B2Granted
Unsupervised clustering feature engineering
AssigneeEarly Warning Services, LLC
InventorJack Kohoutek
Life of this patent
Estimating the term…
Abstract
A method of generating an input for a machine learning algorithm may include collecting data records. Each data record may include a plurality of categories of data. The method may include using vector quantization to partition the plurality of data records into a plurality of groupings. Each of the groupings may be based on one or more of the plurality of categories of data. The method may include generating a correlation score for each of the plurality of groupings. The correlation score may be indicative of whether a particular group is indicative of a given outcome.
Claims21
A method of generating an input for a machine learning algorithm, using one or more processors, comprising:
- collecting a plurality of data records, wherein each data record comprises a plurality of categories of data associated with a particular transaction;
- executing an unsupervised vector quantization machine learning algorithm a plurality of times to partition the plurality of data records into a plurality of first machine learning inputs, wherein:
- each execution of the unsupervised vector quantization machine learning algorithm generates a grouping of data comprising a plurality of subgroups;
- each grouping of data comprises a unique combination of the one or more of the plurality of categories of data;
- each of the plurality of subgroups within each grouping of data comprises one or more data records of the plurality of data records that have values that fall within a predetermined range for the unique combination of the one or more of the plurality of categories of data of the respective grouping of data;
- and
- each of the plurality of first machine learning inputs comprises one subgroup of the plurality of subgroups;
- generating a second machine learning input for each grouping of data, wherein:
- the second machine learning input comprises a correlation score that is indicative of whether a particular subgroup within the groupings of data is indicative of whether the particular transaction is fraudulent;
- and
- each correlation score is based at least partially on whether any transactions within a particular subgroup of data of a grouping of data associated with the correlation score are fraudulent;
- and
- providing, based on the correlation score, at least one of the plurality of first machine learning inputs and associated second machine learning input to a computing device storing the machine learning algorithm that identifies whether a particular data record is likely to be associated with a fraudulent transaction.
10 more elements
A system, comprising:
- one or more computing devices;
- and
- memory storing instructions, the instructions being executable by the one or more computing devices, wherein the one or more computing devices are configured to:
- collect a plurality of data records, wherein each data record comprises a plurality of categories of data associated with a particular transaction;
- executing an unsupervised vector quantization machine learning algorithm a plurality of times to partition the plurality of data records into a plurality of first machine learning inputs, wherein:
- each execution of the unsupervised vector quantization machine learning algorithm generates a grouping of data comprising a plurality of subgroups;
- each grouping of data comprises a unique combination of the one or more of the plurality of categories of data;
- each of the plurality of subgroups within each grouping of data comprises one or more data records of the plurality of data records that have values that fall within a predetermined range for the unique combination of the one or more of the plurality of categories of data of the respective grouping of data;
- and
- each of the plurality of first machine learning inputs comprises one subgroup of the plurality of subgroups;
- generating a second machine learning input for each grouping of data, wherein:
- the second machine learning input comprises a correlation score that is indicative of whether a particular subgroup within the groupings of data is indicative of whether the particular transaction is fraudulent;
- and
- each correlation score is based at least partially on whether any transactions within a particular subgroup of data of a grouping of data associated with the correlation score are fraudulent;
- and
- providing, based on the correlation score, at least one of the plurality of first machine learning inputs and associated second machine learning input to a computing device storing a machine learning algorithm that identifies whether a particular data record is likely to be associated with a fraudulent transaction.
13 more elements
A non-transitory computing-device readable storage medium on which computing-device readable instructions of a program are stored, the instructions, when executed by one or more computing devices, causing the one or more computing devices to perform a method, comprising:
- collecting a plurality of data records, wherein each data record comprises a plurality of categories of data associated with a particular transaction;
- executing an unsupervised vector quantization machine learning algorithm a plurality of times to partition the plurality of data records into a plurality of first machine learning inputs, wherein:
- each execution of the unsupervised vector quantization machine learning algorithm generates a grouping of data comprising a plurality of subgroups;
- each grouping of data comprises a unique combination of the one or more of the plurality of categories of data;
- each of the plurality of subgroups within each grouping of data comprises one or more data records of the plurality of data records that have values that fall within a predetermined range for the unique combination of the one or more of the plurality of categories of data of the respective grouping of data;
- and
- each of the first plurality of machine learning inputs comprises one subgroup of the plurality of subgroups;
- generating a second machine learning input for each grouping of data, wherein:
- the second machine learning input comprises a correlation score that is indicative of whether a particular subgroup within the groupings of data is indicative of whether the particular transaction is fraudulent;
- and
- each correlation score is based at least partially on whether any transactions within a particular subgroup of data of a grouping of data associated with the correlation score are fraudulent;
- and
- providing, based on the correlation score, at least one of the plurality of first machine learning inputs and associated second machine learning input to a computing device storing a machine learning algorithm that identifies whether a particular data record is likely to be associated with a fraudulent transaction.
10 more elements
Dependencies are inferred from the claim text.