Data classifier for classifying pattern data into clusters

• US 20030169919A1
• Filed: 02/28/2003
• Published: 09/11/2003
• Est. Priority Date: 03/05/2002
• Status: Active Grant

• Current Assignee: Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Original Assignee: Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Inventors: Ikeda, Hitoshi, Kato, Noriji, Kashimura, Hirotsugu
• Granted Patent: US 7,227,985 B2
• Time in Patent Office: Days
• US Class Current: 382/159
• CPC Class Codes: G06F 18/23 Clustering techniques

Abstract

A data classifier classifies a plurality of input pattern data into one or more clusters. For each pattern data, a cluster to which the pattern data belongs is provisionally determined. For each cluster, a predetermined correlation value is calculated between one or more pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster. A cluster to which the observational pattern data belongs is determined based on the correlation values.

Claims

1. A data classifier for classifying a plurality of input pattern data into one or more clusters, wherein

initially, for each pattern data, a cluster to which the pattern data belongs is provisionally determined;

for each cluster, a predetermined correlation value is calculated between one or more pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster; and

a cluster to which the observational pattern data belongs is determined based on the correlation value.

2. A data classifier for classifying a plurality of input pattern data into one or more clusters, wherein

initially, for each pattern data, a cluster to which the pattern data belongs is provisionally determined;

for each cluster, each degree of similarity between each pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster is determined using a predetermined function and a correlation value is calculated by summing the values of each degree of similarity; and

a cluster to which the observational pattern data belongs is determined based on the correlation value determined for each cluster.

3. A data classifier according to claim 2, wherein

the predetermined function is defined including a parameter based on an occurrence frequency distribution of a degree of similarity of pattern data or a parameter based on a statistical value of the values of the degree of similarity for each pattern data with other pattern data.

4. A data classifier according to claim 2, wherein

the predetermined function is a function wherein the value relatively rapidly and asymptotically approaches a minimum value with a decrease in the degree of similarity.

5. A data classifier for classifying a plurality of input pattern data into one or more clusters, wherein

(a) for each pattern data, a cluster to which the pattern data belongs is provisionally determined;

(b) each pattern data is sequentially selected as observational pattern data which is to become a target to be classified into a cluster;

(c) for each cluster, a predetermined correlation value is calculated between one or more pattern data belonging to the cluster and the observational pattern data which is a target to be classified into a cluster;

(d) a cluster to which the observational pattern data should belong is determined based on the correlation value; and

the processes of (b), (c), and (d) are repeated until there is no change in the cluster to which each pattern data should belong, and each pattern data is classified into a cluster.

6. A data classification method for classifying a plurality of input pattern data into one or more clusters, comprising the steps of:

provisionally determining, for each pattern data, a cluster to which the pattern data belongs;

calculating, for each cluster, a predetermined correlation value between one or more pattern data belonging to the cluster and observational pattern data which is a target to be classified; and

determining a cluster to which the observational pattern data belongs based on the correlation value.

7. A data classification method for classifying a plurality of input pattern data into one or more clusters, comprising the steps of:

provisionally determining, for each pattern data, a cluster to which the pattern data belongs;

for each cluster, determining, using a predetermined function, a degree of similarity between each pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster and calculating a correlation value by summing the values of the degree of similarity; and

determining a cluster to which the observational pattern data belongs based on the correlation value for each cluster.

8. A data classification method according to claim 7, wherein

the predetermined function is defined including a parameter based on an occurrence frequency distribution of a degree of similarity of pattern data or a parameter based on a statistical value of the values of the degree of similarity for each pattern data with other pattern data.

9. A data classification method according to claim 7, wherein

the predetermined function is a function wherein the value relatively rapidly and asymptotically approaches a minimum value with a decrease in the degree of similarity.

10. A data classification program which, when executed, causes a computer to classify a plurality of input pattern data into one or more clusters by executing the steps of:

provisionally determining, for each pattern data, a cluster to which the pattern data belongs;

calculating, for each cluster, a predetermined correlation value between one or more pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster; and

determining a cluster to which the observational pattern data belongs based on the correlation value.

11. A data classification program which, when executed, causes a computer to classify a plurality of input pattern data into one or more clusters by executing the steps of:

provisionally determining, for each pattern data, a cluster to which the pattern data belongs;

for each cluster, determining, using a predetermined function, a degree of similarity between each pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster and calculating a correlation value by summing the values of the degree of similarity; and

determining a cluster to which the observational pattern data belongs based on the correlation value of each cluster.

12. A data classification program according to claim 11, wherein

the predetermined function is defined including a parameter based on an occurrence frequency distribution of a degree of similarity of pattern data or a parameter based on a statistical value of the values of the degree of similarity for each pattern data with other pattern data.

13. A data classification program according to claim 11, wherein

the predetermined function is a function wherein the value relatively rapidly and asymptotically approaches a minimum value with a decrease in the degree of similarity.