System to automatically categorize
AssigneeAmerican Express Travel Related Services Company, Inc.
InventorsRobert Frohwein, Benjamin Cowen
Life of this patent
Estimating the term…
Abstract
A machine-learning system which utilizes neural networks and other machine learning concepts to automatically categorize a stream of data. Known financial data can be analyzed and trained as part of a machine learning system. Then, a new set of a financial data for a particular user can be analyzed and automatic determinations can be made whether the user is a small business or consumer, and what industry the user is in.
Claims20
A method, comprising:
- processing, by at least one electronic processor, financial transaction data stored in a database into features for a plurality of vendor classifications, wherein the processing comprises generating a set of feature vectors for each vendor classification among the plurality of vendor classifications via execution of a local feature selection algorithm;
- training a neural network utilizing the features for the plurality of vendor classifications;
- retrieving the processed financial transaction data for a particular user;
- identifying classifications of the particular user, wherein the identifying is performed by the neural network;
- displaying the classifications of the particular user and receiving feedback from the particular user;
- and
- further training the neural network using the feedback.
An apparatus, comprising:
- at least one non-transitory computer readable storage medium storing computer readable instructions;
- at least one electronic processor configured to read and execute the computer readable instructions from the at least one non-transitory computer readable storage medium, the computer readable instructions configured to, when executed, cause the at least one electronic processor to:
- access a database of financial transaction data for a plurality of users;
- process the financial transaction data into features for a plurality of vendor classifications, wherein the processing comprises generating a set of feature vectors for each vendor classification among the plurality of vendor classifications via execution of a local feature selection algorithm;
- train a neural network utilizing the features for the plurality of vendor classifications;
- retrieve financial transaction data for a particular user;
- identify classifications of the particular user, the identify is performed by the neural network;
- display the classifications of the particular user and receiving feedback from the particular user;
- and
- further training the neural network using the feedback.
A non-transitory computer readable medium storing program instructions, which upon execution of the program instructions by one or more electronic processors, causes the one or more electronic processors to perform the operations of:
- accessing a database of financial transaction data for a plurality of users;
- processing the financial transaction data into features for a plurality of vendor classifications, wherein the processing comprises generating a set of feature vectors for each vendor classification among the plurality of vendor classifications via execution of a local feature selection algorithm;
- training a neural network utilizing the features for the plurality of vendor classifications;
- retrieving financial transaction data for a particular user;
- identifying classifications of the particular user, the identifying is performed by the neural network;
- displaying the classifications of the particular user and receiving feedback from the particular user;
- and
- further training the neural network using the feedback.
Dependencies are inferred from the claim text.