Apparatus for data decomposition and method and storage medium therefor
First Claim
1. An apparatus for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
- a unit figuring evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events; and
a unit extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein the evaluation values are figured by including a selected event count, a total segment count in partial feature spaces that are to be selected, and a selected segment count in the partial feature spaces that are to be selected.
1 Assignment
0 Petitions
Accused Products
Abstract
The feature quantity selection vectors, indicating which features among those respecting discrete events are to be selected, are fixed. The feature quantity selection spaces taking discrete features as coordinate axes are divided into each segment, and the number of events contained therein is counted. The segments are arranged in order of diminishing event counts contained therein, and the values for functions evaluating the extent of relevance are scanned in the order of segments for which the segments are arranged from the maximum segment counts in the descending order. When the partial situations determined by selection segment counts that are maximum values take evaluation values as their maximum values, even where the feature quantity selection counts are changed, those partial situations are extracted.
62 Citations
36 Claims
-
1. An apparatus for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
-
a unit figuring evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events; and
a unit extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein the evaluation values are figured by including a selected event count, a total segment count in partial feature spaces that are to be selected, and a selected segment count in the partial feature spaces that are to be selected. - View Dependent Claims (2, 3)
said apparatus for data decomposition utilizes an evaluation function which increase monotonically with respect to the number of events selected, increases monotonically with respect to an all segment count in the partial feature spaces that are to be selected, and decreases monotonically with respect to a selected segment count in the partial feature spaces that are to be selected, and said apparatus for data decomposition extracts partial data corresponding to the maximum value for an evaluation value. -
3. The apparatus for data decomposition according to claim 2, wherein the partial data corresponding to a minimum value are extracted utilizing an evaluation function in which increase/decrease relationship are inverted compared with said evaluation function.
-
-
4. An apparatus for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
-
a unit figuring evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events; and
a unit extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein an evaluation value is figured using parameters comprising ratios of event counts included in the partial situations which are the partial data among whole data and are designated by selected events and selected features, average values for the selected event counts for each segment in the partial feature spaces, and space occupation ratios for the selected segment counts in the partial feature spaces, and wherein said apparatus for data decomposition utilizes an evaluation function which increases monotonically with respect to the ratio of event counts included in a partial situation, increases monotonically with respect to the average value of selected event counts for each segment in the partial feature spaces, and decreases monotonically with respect to the space occupation ratio of selected segment counts in the partial feature spaces, and said apparatus for data decomposition extracts the partial data corresponding to the maximum value for an evaluation value which is an output value of said evaluation function. - View Dependent Claims (5, 6, 7)
the partial data corresponding to a minimum value are figured using an evaluation function in which increase/decrease relationships are inverted compared with said evaluation function. -
6. The apparatus for data decomposition according to claim 4, wherein
said apparatus for data decomposition uses C1, C2, C3 as positive constants, n/N as a ratio of the number of events included in a partial situation, n/rd as an average value of the selected event counts for each segment in the partial feature spaces, rd/Sd as an occupation ratio for the selected segment counts in the partial feature spaces, and -
( n , r d _ , N ) = C 1 log n N + C 2 log n r d _ - C 3 log r d _ S d _ as said evaluation function.
-
-
7. The apparatus for data decomposition according to claim 6, wherein
C1, C2, C3 are set so that C1=k, C2=0.3k, and C3=0.7k (k is a constant).
-
-
8. An apparatus for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
-
a unit figuring evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events;
a unit extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events;
a unit generating combinations of feature selections;
a unit searching for the extremal values of evaluation values according to changes in the combinations of events in specific partial feature spaces generated by the generating unit;
a unit confirming that the combinations of events that will become extremal values are extremal values with respect to minute changes in feature quantity selection directions; and
a unit outputting results of the event and feature selections, which are extremal values with respect also to any change from said searching unit and said confirmation unit. - View Dependent Claims (9, 10, 11, 12)
the segments comprising no events are not selected in segment selection. -
11. The apparatus for data decomposition according to claim 9, wherein
all events contained in the selected segments are selected. -
12. The apparatus for data decomposition according to claim 9, wherein
the segments with high event counts are selected preferentially.
-
-
13. A method for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
-
calculating evaluation values that become standards for ascertaining the relevance among data with respect to combinations of specific features and combinations of specific events, and extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein the evaluation values are figured by including a selected event count, a total segment count in partial feature spaces that are to be selected, and a selected segment count in the partial feature spaces that are to be selected. - View Dependent Claims (14, 15)
said method for data decomposition utilizes an evaluation function which increase monotonically with respect to a number of events selected, increases monotonically with respect to an all segment count in the partial feature spaces that are selected, and decreases monotonically with respect to a selected segment count in the partial feature spaces that are selected, and said method for data decomposition extracts the partial data corresponding to the maximum value for the evaluation value. -
15. The method for data decomposition according to claim 14, wherein the partial data corresponding to a minimum value are extracted utilizing an evaluation function in which increase/decrease relationships are inverted compared with said evaluation function.
-
-
16. A method for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
-
calculating evaluation values that become standards for ascertaining the relevance among data with respect to combinations of specific features and combinations of specific events, and extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein an evaluation value is figured using parameters comprising ratios of event counts included in partial situations which are partial data among whole data and are designated by selected events and selected features quantities, average values for selected event counts for each segment in the partial feature spaces, and space occupation ratios for selected segment counts in the partial feature spaces, wherein said method for data decomposition utilizes an evaluation function which increases monotonically with respect to the ratio of event counts included in a partial situation, increases monotonically with respect to the average value of selected event counts for each segment in the partial feature spaces, and decreases monotonically with respect to the space occupation ratio of selected segment counts in the partial feature spaces, and said method for data decomposition extracts the partial data corresponding to a maximum value for an evaluation value which is an output value of said evaluation function. - View Dependent Claims (17, 18, 19)
the partial data corresponding to a minimum value are figured using an evaluation function in which increase/decrease relationships are inverted compared with said evaluation function. -
18. The method for data decomposition according to claim 16, wherein
said method for data decomposition uses C1, C2, C3 as positive constants, n/N as a ratio of the number of events included in a partial situation, n/rd as an average value of the selected event counts for each segment in the partial feature spaces, rd/Sd as an occupation ratio for selected segment counts in the partial feature spaces, and -
( n , r d _ , S d _ , N ) = C 1 log n N + C 2 log n r d _ - C 3 log r d _ S d _ as said evaluation function.
-
-
19. The method for data decomposition according to claim 18, wherein
C1, C2, C3 are set so that C1=k, C2=0.3k, and C3=0.7k (k is a constant).
-
-
20. A method for data decomposition that extracts partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which are said attributes, comprising:
-
calculating evaluation values that become standards for ascertaining the relevance among data with respect to combinations of specific features and combinations of specific events, extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, generating the combinations of feature selections;
searching for the extremal values of evaluation values according to changes in the combinations of events in specific partial feature spaces generated by the generating;
confirming that the combinations of events that will become extremal values are extremal values with respect to minute changes in the feature selection directions; and
outputting results of event and feature selections, which are extremal values with respect also to any change from said searching and said confirming. - View Dependent Claims (21, 22, 23, 24)
the segments comprising no events are not selected in segment selection. -
23. The method for data decomposition according to claim 21, wherein
all events contained in the selected segments are selected. -
24. The method for data decomposition according to claim 21, wherein
the segments with high event counts are selected preferentially.
-
-
25. A computer-readable storage medium for extracting partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which comprise the attributes, causing a computer to perform a process comprising:
-
calculating evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events; and
extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein the evaluation values are figured by including a selected event count, a total segment count in partial feature spaces that are to be selected, and a selected segment count in the partial feature spaces that are to be selected. - View Dependent Claims (26, 27)
said storage medium makes the computer extract the partial data corresponding to the maximum value for an evaluation value. -
27. The storage medium according to claim 26, wherein the partial data corresponding to a minimum value are extracted utilizing an evaluation function in which increase/decrease relationships are inverted compared with said evaluation function.
-
-
28. A computer-readable storage medium for extracting partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which comprise the attributes, causing a computer to perform a process comprising:
-
calculating evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events; and
extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events, wherein an evaluation value is figured using parameters comprising ratios of event counts included in partial situations which are partial data among whole data and are designated by selected events and selected features, average values for selected event counts for each segment in the partial feature spaces, and space occupation ratios for selected segment counts in the partial feature spaces, and wherein said storage medium makes the computer utilize an evaluation function which increases monotonically with respect to the ratio of event counts included in a partial situation, increases monotonically with respect to the average value of selected event counts for each segment in the partial feature spaces, and decreases monotonically with respect to the space occupation ratio of selected segment counts in the partial feature spaces, and said storage medium makes the computer extract partial data corresponding to the maximum value for an evaluation value which is an output value of said evaluation function. - View Dependent Claims (29, 30, 31)
the partial data corresponding to a minimum value are figured using an evaluation function in which increase/decrease relationships are inverted compared with said evaluation function. -
30. The storage medium according to claim 28, wherein
said storage medium makes the computer use C1, C2, C3 as positive constants, n/N as a ratio of the number of events included in a partial situation, n/rd as an average value of selected event counts for each segment in the partial feature spaces, rd/Sd as an occupation ratio for selected segment counts in the partial feature spaces, and -
( n , r d _ , S d _ , N ) = C 1 log n N + C 2 log n r d _ - C 3 log r d _ S d _ as said evaluation function.
-
-
31. The storage medium according to claim 30, wherein
C1, C2, C3 are set so that C1=k, C2=0.3k, and C3=0.7k (k is a constant).
-
-
32. A computer-readable storage medium for extracting partial data from whole data by selecting, with respect to each record and from data which catalog a plurality of attributes possessed by each record, a combination of each event corresponding to each record and a combination of feature quantities, which comprise the attributes, causing a computer to perform a process comprising:
-
calculating evaluation values that become standards for ascertaining a relevance among data with respect to combinations of specific features and combinations of specific events;
extracting a plurality of partial data for which an evaluation value becomes a maximum value with respect to both changes in combinations of features and changes in combinations of events;
generating combinations of feature quantity selections;
searching for extremal values of evaluation values according to changes in the combinations of events in specific partial feature spaces generated by the generating function;
confirming that the combinations of events that will become extremal values are extremal values with respect to minute changes in the feature selection directions; and
outputting results of event and feature selections, which are extremal values with respect also to any change from said searching function and said confirming function. - View Dependent Claims (33, 34, 35, 36)
the segments comprising no events are not selected in segment selection. -
35. The storage medium according to claim 33, wherein
all events contained in the selected segments are selected. -
36. The storage medium according to claim 33, wherein
the segments with high event counts are selected preferentially.
-
Specification