Prediction analysis apparatus and program storage medium therefor

• US 20020049685A1
• Filed: 03/27/2001
• Published: 04/25/2002
• Est. Priority Date: 10/25/2000
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Yaginuma, Yoshinori
• Granted Patent: US 6,988,090 B2
• Time in Patent Office: Days
• US Class Current: 706/21
• CPC Class Codes: G06N 3/02 Neural networks

Abstract

A prediction analysis apparatus includes a prediction unit for predicting a result value corresponding to one or more attribute values of unknown data using known data indicating the correspondence between one or more attribute values and corresponding result values, and an analysis unit for outputting analysis information indicating how at least one attribute value of the unknown data is to be amended to change a result value predicted by the prediction unit into a desired prediction value. The prediction analysis apparatus with the configuration can be realized by, for example, a computer executing a program for realizing the function of each of the above mentioned units.

Claims

1. A prediction analysis apparatus, comprising:

a prediction unit predicting a result value corresponding to one or more attribute values of unknown data using known data indicating correspondence between one or more attribute values and corresponding result values; and

an analysis unit outputting analysis information indicating how at least one attribute value of the unknown data is to be amended to change a result value predicted by said prediction unit into a desired prediction value.

2. The apparatus according to claim 1, wherein

said analysis unit extracts known data having the desired prediction value as a result value, and having one or more attribute values similar to one or more attribute values of the unknown data from know data, and outputs the extracted known data as the analysis information.

3. The apparatus according to claim 2, wherein

said analysis unit extracts known data similar to the unknown data from the known data with an importance factor of each attribute taken into account.

4. The apparatus according to claim 3, wherein

said analysis unit uses an influence factor on a result value from each attribute obtained by memory-based reasoning as the importance factor.

5. The apparatus according to claim 3, wherein

said analysis unit uses a weight obtained from learning of a structured neural network as the importance factor.

6. The apparatus according to claim 2, wherein

said analysis unit generates one piece of known data by performing a predetermined operation on plural pieces of known data when the plural pieces of known data are extracted from the known data, and outputs the generated known data.

7. The apparatus according to claim 2, wherein

said analysis unit outputs predetermined pieces of known data in order from data most similar to the unknown data in plural pieces of known data when the plural pieces of known data are extracted from the known data.

8. The apparatus according to claim 1, wherein

said analysis unit outputs at least one attribute value of unknown data whose desired prediction value is to be predicted as a result value, or an amount of a change into the at least one attribute value as the analysis information.

9. The apparatus according to claim 8, wherein

said analysis unit specifies the at least one attribute value of unknown data whose desired prediction value is to be predicted as a result value, or an amount of a change into the at least one attribute value through a neural network.

10. The apparatus according to claim 1, wherein

said analysis unit refers to a decision tree, specifies a path through which the desired prediction value is to be predicted as a result value of the unknown data, extracts known data whose result value is predicted through the specified path, and outputs the extracted known data as the analysis information.

11. The apparatus according to claim 10, wherein

said analysis unit also outputs a certainty factor in the specified path.

12. The apparatus according to claim 1, wherein

when a result value of unknown data is predicted by referring to rules indicating one or more condition units presenting a condition of the attribute value and a result value under a condition indicated by the condition units, said analysis unit changes a condition indicated by a condition unit in the condition units of a rule used to predict a result value of the unknown data in the rules so that a rule to be referenced in the rules for prediction of the desired prediction value as a result value of the unknown data, known data whose result value can be predicted based on the specified rule and which has a desired prediction value as the result value can be extracted, and the extracted known data can be output as the analysis information.

13. The apparatus according to claim 1, wherein

said analysis unit sets an attribute whose attribute value is to be changed in attributes of the unknown data, and obtains the analysis information by changing the attribute value of the set attribute.

14. The apparatus according to claim 13, wherein

said attribute to be changed can be set by a user in an interactive mode.

15. The apparatus according to claim 13, wherein

said analysis unit sets the attribute to be changed with an importance factor of each attribute taken into account.

16. The apparatus according to claim 15, wherein

said analysis unit uses an influence factor on a result value from each attribute obtained by memory-based reasoning as the importance factor.

17. The apparatus according to claim 15, wherein

said analysis unit uses a weight obtained from learning of a structured neural network as the importance factor.

18. The apparatus according to claim 13, wherein

said analysis unit sets a search range of an attribute value of an attribute set to be changed, and obtains the analysis information by changing an attribute value of the attribute set to be changed in a corresponding search range.

19. A prediction analysis apparatus, comprising:

a prediction unit predicting a result value corresponding to one or more attribute values of unknown data according to predicting information for predicting the result value; and

an analysis unit outputting analysis information indicating how at least one attribute value of the unknown data is to be amended to change a result value predicted by said prediction unit into a desired prediction value.

20. A computer-readable storage medium storing a program used to direct a computer to perform the processes of:

predicting a result value corresponding to one or more attribute values of unknown data using known data indicating correspondence between one or more attribute values and corresponding result values; and

outputting analysis information indicating how at least one attribute value of the unknown data is to be amended to change a result value predicted in said predicting process into a desired prediction value.