INFORMATION CLASSIFICATION SYSTEM

• US 20130091137A1
• Filed: 06/15/2011
• Published: 04/11/2013
• Est. Priority Date: 06/25/2010
• Status: Active Grant

• Current Assignee: NEC Communication Systems Limited (NEC Corporation)
• Original Assignee: NEC Communication Systems Limited (NEC Corporation)
• Inventors: Aust, Stefan, Matsumoto, Akira, Ito, Tetsuya, Yoshimura, Koichi, Kuroshima, Soshi
• Granted Patent: US 9,009,160 B2
• Time in Patent Office: Days
• US Class Current: 707/737
• CPC Class Codes: G06F 16/35 Clustering; Classification

Abstract

An information classification system 100 includes: a classified information storing part 101 for storing classified information having already been classified into a certain group and group specification information for specifying the group, in association with each other; and a confidence calculating part 102 for calculating confidence having a value depending on a probability that a group which is specified based on unclassified information as a target to be classified and classified information selected as reference information from among the stored classified information and into which the unclassified information should be classified is a true group.

Claims

1. An information classification system, comprising:

a classified information storing unit for storing classified information having already been classified into a certain group and group specification information for specifying the group, in association with each other; and

a confidence calculating unit for calculating confidence having a value depending on a probability that a group which is specified based on unclassified information as a target to be classified and classified information selected as reference information from among the stored classified information and into which the unclassified information should be classified is a true group.

2. The information classification system according to claim 1, wherein the confidence calculating unit is configured to:

with respect to each classified information of the stored classified information, judge whether a first group and a second group match each other, the first group being a group which is specified based on the each classified information as judgment target information and based on classified information selected as a reference information candidate from among the stored classified information and into which the judgment target information should be classified, and the second group being a group specified based on the group specification information stored in association with the judgment target information; and

calculate static confidence as one of the confidence based on a matching number as a number of judgment target information that the first group and the second group match each other.

3. The information classification system according to claim 2, wherein the confidence calculating unit is configured to calculate the static confidence having a value becoming larger as a ratio of the matching number to a number of the stored classified information becomes higher.

4. The information classification system according to claim 2, wherein the confidence calculating unit is configured to:

with respect to each classified information of a plurality of classified information stored in association with same group specification information, acquire the matching number by selecting the each classified information as the reference information candidate; and

from among the plurality of classified information stored in association with the same group specification information, select classified information that the acquired matching number is maximum as the reference information for the group specification information.

5. The information classification system according to claim 1, wherein the confidence calculating unit moans is configured to:

with respect to each group specification information of group specification information different from each other among the stored group specification information, calculate a divergence representing a degree of divergence between the reference information selected for the each group specification information and the unclassified information; and

calculate dynamic confidence as one of the confidence based on the calculated divergence.

6. The information classification system according to claim 5, wherein the confidence calculating unit is configured to calculate the dynamic confidence having a value becoming smaller as a difference between a minimum divergence that is a minimum value of the calculated divergence and a proximate divergence that is a second minimum value next to the minimum divergence of the calculated divergence becomes smaller.

7. The information classification system according to claim 1, comprising:

an unclassified information acquiring unit for acquiring the unclassified information; and

a group specifying unit for, based on the acquired unclassified information and the reference information, specifying a group into which the unclassified information should be classified.

11. The information classification system according to claim 5, wherein the confidence calculating unit is configured to calculate the divergence by using Kullback-Leibler divergence (KL divergence) or Jeffrey divergence.

12. The information classification system according to claim 1, wherein:

the classified information is communication status information representing a status of wireless communication having been executed by a communication terminal;

the unclassified information is communication status information representing a status of wireless communication being currently executed by a communication terminal; and

the group specification information is location specification information for specifying a location of a communication terminal.

17. The information classification system according to claim 3, wherein the confidence calculating unit is configured to:

with respect to each classified information of a plurality of classified information stored in association with same group specification information, acquire the matching number by selecting the each classified information as the reference information candidate; and

from among the plurality of classified information stored in association with the same group specification information, select classified information that the acquired matching number is maximum as the reference information for the group specification information.

18. The information classification system according to claim 2, wherein the confidence calculating unit is configured to:

with respect to each group specification information of group specification information different from each other among the stored group specification information, calculate a divergence representing a degree of divergence between the reference information selected for the each group specification information and the unclassified information; and

calculate dynamic confidence as one of the confidence based on the calculated divergence.

19. The information classification system according to claim 3, wherein the confidence calculating unit is configured to:

with respect to each group specification information of group specification information different from each other among the stored group specification information, calculate a divergence representing a degree of divergence between the reference information selected for the each group specification information and the unclassified information; and

calculate dynamic confidence as one of the confidence based on the calculated divergence.

20. The information classification system according to claim 4, wherein the confidence calculating unit is configured to:

with respect to each group specification information of group specification information different from each other among the stored group specification information, calculate a divergence representing a degree of divergence between the reference information selected for the each group specification information and the unclassified information; and

calculate dynamic confidence as one of the confidence based on the calculated divergence.

8. An information classification method, comprising calculating confidence having a value depending on a probability that a group into which unclassified information as a target to be classified should be classified is a true group, the group being specified based on the unclassified information and based on classified information selected as reference information from among classified information stored in a storage device for storing classified information having already been classified into a certain group and group specification information for specifying the group in association with each other.

9. An information classification device, comprising:

a classified information storing unit for storing classified information having already been classified into a certain group and group specification information for specifying the group, in association with each other; and

a confidence calculating unit for calculating confidence having a value depending on a probability that a group which is specified based on unclassified information as a target to be classified and classified information selected as reference information from among the stored classified information and into which the unclassified information should be classified is a true group.

16. The information classification device according to claim 9, wherein the confidence calculating unit is configured to:

with respect to each group specification information of group specification information different from each other among the stored group specification information, calculate a divergence representing a degree of divergence between the reference information selected for the each group specification information and the unclassified information; and

calculate dynamic confidence as one of the confidence based on the calculated divergence.

10. (canceled)

13. The information classification method according to Supplementary Note 8, comprising:

with respect to each classified information of the stored classified information, judging whether a first group and a second group match each other, the first group being a group which is specified based on the each classified information as judgment target information and based on classified information selected as a reference information candidate from among the stored classified information and into which the judgment target information should be classified, and the second group being a group specified based on the group specification information stored in association with the judgment target information; and

calculating static confidence as one of the confidence based on a matching number as a number of judgment target information that the first group and the second group match each other.

14. The information classification method according to Supplementary Note 8, comprising:

with respect to each group specification information of group specification information different from each other among the stored group specification information, calculating a divergence representing a degree of divergence between the reference information selected for the each group specification information and the unclassified information; and

calculating dynamic confidence as one of the confidence based on the calculated divergence.

15. The information classification device according to Supplementary Note 9, wherein the confidence calculating unit is configured to:

with respect to each classified information of the stored classified information, judge whether a first group and a second group match each other, the first group being a group which is specified based on the each classified information as judgment target information and based on classified information selected as a reference information candidate from among the stored classified information and into which the judgment target information should be classified, and the second group being a group specified based on the group specification information stored in association with the judgment target information; and

calculate static confidence as one of the confidence based on a matching number as a number of judgment target information that the first group and the second group match each other.