METHOD AND SYSTEM FOR-CLUSTERING DATA ARISING FROM A DATABASE

• US 20120109959A1
• Filed: 12/21/2009
• Published: 05/03/2012
• Est. Priority Date: 12/30/2008
• Status: Active Grant

• Current Assignee: Thales SA
• Original Assignee: Thales SA
• Inventors: Benhadda, Hamid
• Granted Patent: US 8,458,182 B2
• Time in Patent Office: Days
• US Class Current: 707/737
• CPC Class Codes: G06F 16/285 Clustering or classification

Abstract

A method for clustering data or objects taking the form of an array, each of the elements of said array corresponding to a value of similarity existing between said objects is implemented within a computer linked with a database containing the data or objects to be classified comprising a work memory, and a processor. The method includes steps to determine a number of classes of objects by taking account of the values of the relationships computed between an object and a previously established previous class, for each of the classes found, determine the value of each of the relationships between a class and the other classes, and merge certain classes, and take each object of each class one by one, determine the value of the relationship-of each object with each of the classes other than the class into which it was classed in the initial step, if the value of the relationship is greater then transfer the object from its class to the new class, this is continued until all the values of the relationships are negative.

Claims

1. A method for clustering data or objects O_i, O_j taking the form of an array S, each of the elements of said array S corresponding to a value of similarity s_ij existing between said objects O_i, O_j said method being implemented within a computer linked with a database containing the data or objects to be clustered comprising a work memory, and a processor, the method comprising:

1) determining a number k of classes of objects in the following manner;

take an object O_i which has the fewest possible similarities s_ij with the other objects O_j that is to say the one for which the sum of the components

2. The method according to claim 1, wherein a square array representative of n×

n objects is used.

3. The method according to claim 1, wherein a rectangular array representative of n objects×

a attributes is used, and a step of transforming the array n objects×

a attributes to the array S containing the values of similarity s_ij existing between said objects O_i is executed.

4. A system to cluster data taking the form of an array S, each of the elements of said array S corresponding to a value of similarity s_ij existing between said objects O_j, the system comprising:

a computer (1) linked with a database to be classified comprising a work memory,

a processor to implement the method according to claim 1.

5. The system according to claim 4, further comprising:

a database containing the result of the classification method according to the invention, executed on the processor.

6. The system according to claim 4, further comprising:

inputs/outputs or a man machine interface.