Data processing: artificial intelligence – Knowledge processing system – Knowledge representation and reasoning technique
Reexamination Certificate
2001-01-08
2004-08-24
Starks, Jr., Wilbert L. (Department: 2121)
Data processing: artificial intelligence
Knowledge processing system
Knowledge representation and reasoning technique
C706S045000, C706S048000
Reexamination Certificate
active
06782376
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates reasoning, prediction method, and a system having a reasoning means built therein which uses data involved in a natural phenomenon, such a social phenomenon as vital statistics, such an economic phenomenon as stock price fluctuation, or such chemical and physical phenomenon as industrial plants.
As a technique for effectively utilizing accumulated data, there has been actively studied an analysis technique oriented to a massive amount of data called data mining. The form of data to be extracted varies from application purpose to application purpose, but one of general purposes is reasoning and prediction of new data.
Memory-based reasoning (MBR) is known as a technique for calculating a reasoning result on the basis of similarities from a large amount of accumulated data. The MBR is summarized in a paper entitled “TOWARD MEMORY-BASED REASONING, Communications of the ACM”, Craig Stanfill, David Waltz, December 1986, Vol. 29, Number 12, pp. 1213-1228 (which will be referred to as the prior art 1, hereinafter).
In the prior art 1, case data given in the form of record is regarded as vectors, and a similarity between cases is judged on the basis of the magnitude of a distance between the vectors. Cases having large similarities are extracted as similar cases, and output field values of the similar cases are weighed by the distances between the vectors and averaged to calculate a reasoning result.
Another technique associated with the MBR is described, e.g., in Japanese Patent No. 2,632,117. The patent will be called the prior art 2, hereinafter. In the prior art 2 is arranged so that an input space is divided so that an output error is smaller than a threshold value and constructed so that a space defined by input variables is meshed. A case database for use in reasoning is created by embedding (quantizing) cases into the meshed space. A reasoning result of the reasoning error is obtained with use of distances in the meshed space.
The prior art 1 fails to provide means for determining the number of similar cases employed for a new case, and the determination of the number of similar cases is done by a user. In general, when viewed from a vector space, a distribution of cases is not always uniform. Accordingly, it is not always that, only by merely specifying the number of employed cases, suitable similar cases taking the inter-vector distances into consideration can be collected.
The prior art 2 is directed to a technique for use in inferring typical cases obtained by decimating cases. However, when viewed from the entire cases before decimated, the inter-vector distance possessed by similar cases for a new case is determined by a mesh size regardless of a case distribution. Since the mesh size is uniform, however, it is not always that suitable similar cases can be collected for each mesh when the case distribution is not uniform. Further, in such a case, it becomes difficult to suitably calculate a reasoning result of the reasoning error.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide a means for enabling suitable reasoning even when a case distribution is not uniform, which tends to decrease a reasoning accuracy in the prior art, and also to provide a means for calculating a confidence degree with which the validity of a reasoning result can be judged.
In accordance with the present invention, similar cases are determined by executing a step of calculating a similarity of a case and thereafter utilizing a distribution of predictor field values of cases having high similarities. That is, since the similar cases are determined by taking the peripheral distribution of the new case into consideration, suitable similar cases can be selected for reasoning even when the case distribution is not uniform.
In accordance with the present invention, further, a confidence degree is determined by utilizing a distribution of predicator field values of cases having high similarities. Thus even when the case distribution is not uniform, a suitable confidence degree can be calculated.
REFERENCES:
patent: 5581664 (1996-12-01), Allen et al.
patent: 5701400 (1997-12-01), Amado
patent: 5950182 (1999-09-01), Godbole et al.
patent: 6418425 (2002-07-01), Maeda et al.
patent: 6477538 (2002-11-01), Yaginuma et al.
patent: 6571251 (2003-05-01), Koski et al.
patent: 2632117 (1997-04-01), None
Khoshgoftarr et al., “Predicting Fault-Prone Modules with Case-Based Reasoning”, Proceedings of the 8th International Symposium on Software Reliability Engineering, Nov. 1997, pp. 27-35.*
Nomoto et al., “Similarity Measures Using Fuzzy Thesauri for Case-Based Reasoning”, 1999 IEEE International Conference on Systems, man and Cybernetics, Oct. 1999, vol. 3, pp. 68-72.*
Gebhardt et al., “On a Tool for Possibilistic Reasoning in Relational Structures”, Proceedings of the 5thIEEE International Conference on Fuzzy Systems, Sep. 1996, vol. 2, pp. 1471-1475.*
Holt et al., “Case-Based Reasoning With Spatial Data”, Proceedings of the 2ndNew Zealand International Two-Stream Conference on Artificial Neural Networks and Expert Systems, Nov. 1995, pp. 385-388.*
Sudkamp, T., “Similarity as a Foundation for Possibility”, 9thIEEE International Conference on Fuzzy Systems, May 2000, vol. pp. 735-740.*
Sun et al., “Intergration of Abductive CBR and Deductive CBR”, 10thIEEE International Conference on Fuzzy Systems, Dec. 2001, vol. 3, pp. 1432-1435.*
De et al., “Case-Based Classification Using Fuzziness and Neural Networks”, IEE Colloquium on Knowledge Discovery and Data Mining (Digest No. 1998/310), May 1998, pp. 6/1-6/3.*
Hirota et al., “A Fuzzy Case Based Reasoning System for The Legal Inference”, Proceedings of the 1998 IEEE Intl Conference on Fuzzy Systems, May 1998, vol. 2, pp. 1350-1354.*
Dvir et al., “Matching Attributes in a Fuzzy Case Based Reasoning”, 18thIntl Conference of the North American Fuzzy Information Processing Society, Jun. 1999, pp. 33-36.*
Fathi-Torbaghan et al., “ICARUS: Integrated Rule-Based and Case-Based Reasoning on the Base of Unsharp Symptoms”, IEEE Intl Conference on Systems, Man and Cybernetics, Oct. 1995, vol. 3, pp. 2424-2427.*
Atiquzzaman et al., “Systolic Array Implementation for Fuzzy Evidential Reasoning Systems”, Proceedings of the Intl Conference on Expert Systems for Development, Mar. 1994, pp. 91-96.*
Nakanishi et al., “A Comparison of Direct Fuzzy Reasoning Methods”, Proceedings of the 1995 IEEE Intl Conference on Fuzzy Systems, Mar. 1995, vol. 4, pp. 1927-1933.*
Wang et al., “Mining Approximate Dependency to Answer Null Queries on Similarity-Based Fuzzy Relational Databases”, 9thInternational Conference on Fuzzy Systems, May 2000, vol. 2, pp. 615-620.*
Vasudevan et al., “Fuzzy Logic in Case-Based Reasoning”, Proceedings of the 1stIntl Joint Conference of the North American Fuzzy Information Processing Society Biannual Conference, Dec. 1994, pp. 301-302.*
Baldwin, “Probabilistic, Fuzzy and Evidential Reasoning in FRIL”, IEE Colloquium on Two Decades of Fuzzy Control-Part 2, Ma 1993, pp. 7/1-7/4.*
Racine et al., “Maintaining Unstructured Case Bases”, Proceedings of the 2ndIntl Conference on Case-Based Reasoning, 1997 pp. 553-564.*
Toward Memory-Based Reasoning, Dec. 1986.
Ito Yukiyasu
Morita Toyohisa
Sato Yoshinori
Booker Kelvin
Hitachi , Ltd.
Mattingly Stanger & Malur, P.C.
Starks, Jr. Wilbert L.
LandOfFree
Reasoning method based on similarity of cases does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Reasoning method based on similarity of cases, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Reasoning method based on similarity of cases will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3346083