Method and system for constructing a classifier

• US 10,311,084 B2
• Filed: 10/18/2006
• Issued: 06/04/2019
• Est. Priority Date: 10/22/2005
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bent, Graham Anthony
• Primary Examiner(s): Rifkin, Ben M
• Application Number:
  US11/550,583
• Publication Number:
  US 20070124300A1
• Time in Patent Office: 4,612 Days
• Field of Search: 707 5
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A method and system for constructing a classifier for a set of records to be classified (108) into predicted classes (351-353) are provided. The set of records that are to be classified (108) are clustered into a plurality of clusters. A first classifier (106) is created that classifies records into the plurality of clusters (321-323) and the first classifier (106) is applied to a set of training records (110), each of the training records (331-334) having a predicted class (306). A classifier (107A-C) may then be created for each sub-set of training records (341-343) classed into each of the plurality of clusters and the classifier (107A-C) applied to a sub-set of records to be classified (311-313) formed in the corresponding cluster.

Claims

1. A computer-implemented method for constructing a classifier classifying data records into specific categories, comprising:

a) clustering a set of records that are to be classified into a plurality of clusters of records by applying a clustering process to the set of records that are to be classified;

b) creating a first classifier based upon the clustered set of records that are to be classified, the first classifier enabled to classify records into a plurality of training clusters;

c) applying the first classifier to a set of training records to create a plurality of training clusters, each record in each training cluster of the plurality of training clusters having a corresponding predicted class;

d) creating separate classifiers for each training cluster of the plurality of training clusters; and

,

e) applying one of the created separate classifiers to a corresponding one of the clusters of records amongst the clustered set of records in order to classify the corresponding one of the clusters of records,

wherein the first classifier and also each separate classifier learns one or more sets of relationships that define each of a known set of predicted classes and wherein the computer-implemented method first processes the set of records that are to be classified to determine a context before using the set of training records to classify the set of records based upon the determined context.

2. The method of claim 1, wherein

step c) classifies the set of training records into the plurality of training clusters.

3. The method of claim 1, wherein

each record of the clustered set of records are labeled with a cluster identifier following step a).

4. The method of claim 3, wherein

each record in each training cluster of the plurality of training clusters are further labeled with the cluster identifier at step c).

5. A computer hardware system for constructing a classifier classifying data records into specific categories, comprising:

at least one processor, wherein the at least one processor configured to perform;

clustering a set of records that are to be classified into a plurality of clusters of records by applying a clustering process to the set of records that are to be classified;

creating a first classifier based upon the clustered set of records that are to be classified, the first classifier enabled to classify records into a plurality of training clusters;

applying the first classifier to a set of training records to create a plurality of training clusters, each record in each training cluster of the plurality of training clusters having a corresponding predicted class;

creating separate classifiers for each training cluster the plurality of training clusters; and

,

applying one of the created separate classifiers to a corresponding one of the clusters of records amongst the clustered set of records in order to classify the corresponding one of the clusters of records, and

wherein the at least one processor first processes the set of records that are to be classified to determine a context before using the set of training records to classify the set of records based upon the determined context.

6. The system of claim 5, wherein

the first classifier classifies the set of training records into the plurality of training clusters.

7. The system of claim 5, wherein

upon clustering the set of records, each of the clustered set of records are labeled with a cluster identifier.

8. The system of claim 5, wherein

applying the first classifier to the set of training records further comprises labeling each record in each training cluster of the plurality of training clusters with a cluster identifier.

9. A non-transitory computer readable storage medium for constructing a classifier, comprising computer readable instructions, which, when executed on a computer system, cause the computer system to perform the steps of:

a) clustering a set of records that are to be classified into a plurality of clusters of records by applying a clustering process to the set of records that are to be classified;

b) creating a first classifier based upon the clustered set of records that are to be classified, the first classifier enabled to classify records into a plurality of training clusters;

c) applying the first classifier to a set of training records to create a plurality of training clusters, each record in each training cluster of the plurality of training clusters having a corresponding predicted class;

d) creating separate classifiers for each training cluster of the plurality of training clusters; and

,

e) applying one of the created separate classifiers to a corresponding one of the clusters of records amongst the clustered set of records in order to classify the corresponding one of the clusters of records, wherein the first classifier and also each separate classifier learns one or more sets of relationships that define each of a known set of predicted classes and wherein the computer system first processes the set of records that are to be classified to determine a context before using the set of training records to classify the set of records based upon the determined context.