JP4892868B2 - Inter-set relationship determination program and inter-set relationship determination device - Google Patents

Description

The present invention relates to an inter-set relationship determination program and an inter-set relationship determination apparatus for determining the relationship between sets having numerical data as elements, and more particularly to appropriately finding the relationship between sets having numerical data as elements. The present invention relates to an inter-set relationship determination program and an inter-set relationship determination apparatus .

  In recent years, when companies and organizations are merged or integrated, the issue of how to integrate and interoperate information held by companies and organizations is increasingly important for the proper management and operation of intellectual assets. It is becoming.

  However, when the classification system for classifying information into categories is different, it is difficult to find a category pair including similar information in order to integrate the information. Specifically, due to differences in category names, category granularity, or differences in the structure of classification hierarchies, simply searching for the corresponding category between classification systems is possible simply by comparing the category names. The problem of not being able to occur.

  FIG. 11 is a diagram for explaining a difference in categories between different classification systems. As shown in FIG. 11, in the classification system 1, there are a category “American football” and a category “football” as subordinate categories of the category “Sports”. In the classification system 2, there are a category “Soccer” and a category “Football” as subcategories of the category “Sports”.

  Here, the word “Football” may be used in a broad sense such as “Soccer” or “American football”. Therefore, the category “Football” in the classification system 1 may actually mean “Soccer”, and the category “Football” in the classification system 2 may actually mean “American football”. In such a case, it is not possible to appropriately associate the categories by simply comparing the category names.

  Therefore, a technique is disclosed in which texts are associated with each other by extracting text data from categories in two classification systems, generating feature vectors from the categories, and comparing them, instead of simply comparing the names of the categories. (For example, see Patent Document 1).

JP 2005-63332 A

  However, the conventional techniques represented by the above-mentioned patent documents perform morphological analysis, feature word extraction, and the like on text data to associate categories, and data belonging to the category is numerical data. In such a case, there is a problem that it is difficult to perform the association.

  When the data belonging to the category is numerical data, it is difficult to characterize the category by each numerical value, and it is difficult to find the correspondence between the categories even if each numerical value is simply examined.

  Therefore, when there are multiple categories consisting of a collection of numerical data, how to properly find the relationship between them becomes an important issue for the appropriate management and operation of intellectual assets. ing.

The present invention has been made to solve the above-described problems caused by the prior art, and is a program for determining the relationship between sets, which can appropriately find the relationship between sets whose elements are numerical data, and between sets. An object of the present invention is to provide a relevance determination device .

  In order to solve the above-described problems and achieve the object, the present invention is an inter-group relevance determination program for determining relevance between sets having numerical data as an element, the statistics relating to the distribution of the numerical data A statistic calculation procedure for calculating the statistic, a statistic storage procedure for storing the statistic calculated by the statistic calculation procedure, and determining an association between sets based on the statistic stored in the statistic storage procedure And causing the computer to execute the relevance determination procedure.

  Also, in the present invention according to the above-described invention, the relevance determination procedure determines whether any one of an equivalence relation, an equivalence relation, a sibling relation, or a hierarchy relation is between sets. It is characterized by.

  In the invention described above, the present invention is characterized in that the set is a set related to attributes of a predetermined concept.

  Further, in the present invention, in the above invention, the unit of the numerical data is adjusted, and when the units do not match, the computer further executes a unit adjustment procedure for determining that there is no relationship between the sets, The statistic calculation procedure calculates a statistic related to the distribution of numerical data whose units are adjusted by the unit adjustment procedure when the units match.

  Further, the present invention is characterized in that, in the above-mentioned invention, the relevance determination procedure determines relevance between sets by comparing a distance between distributions of numerical data of the set with a predetermined threshold.

  Further, in the present invention according to the above-described invention, the relevance determination procedure is such that when there is no overlap between the two distributions of the numerical data in the set, the maximum value of the numerical data in the smaller distribution and the value are large. The relationship between sets is determined by examining whether or not the difference in the number of digits from the minimum value of the numerical data in this distribution is greater than 3 digits.

  Further, the present invention is the above invention, wherein the relevance determination procedure performs an identity test on the average value of the numerical data based on the statistic stored in the statistic storage procedure. It is characterized by determining the relevance of.

  Further, the present invention is the above invention, wherein the relevancy determination procedure is such that the significance level related to the test in the case of performing the test of identity related to the average value of the numerical data is within a range of 0.01 to 0.3. It is characterized by setting.

  Further, the present invention is the above invention, wherein the relevance determination procedure performs an identity test related to the dispersion of the numerical data based on the statistic stored in the statistic storage procedure. It is characterized by determining the relevance of.

  Further, the present invention is the above invention, wherein the relevance determination procedure sets a significance level related to a test in the case of performing an identity test related to the variance of the numerical data within a range of 0.01 to 0.3. It is characterized by doing.

  Further, the present invention is an inter-set relevance determination method for determining relevance between sets having numerical data as an element, the statistical amount calculating step for calculating a statistic related to the distribution of the numerical data, and the statistical A statistic storage step for storing the statistic calculated by the quantity calculation step, and a relevance determination step for determining the relevance between the sets based on the statistic stored in the statistic storage step. It is characterized by.

  According to the present invention, the statistics related to the distribution of numerical data are calculated, the calculated statistics are stored, and the relevance between sets is determined based on the stored statistics. It is possible to appropriately find the relationship between the sets having numerical data as elements from the statistics related to.

  In addition, according to the present invention, since it is determined whether any one of the equivalence relation, the equivalence relation, the sibling relation, or the upper and lower hierarchy relation is between the sets, the relation between the sets is determined. The effect that it can analyze in detail is produced.

  In addition, according to the present invention, since the set is a set related to attributes of a predetermined concept, when the class attribute in the ontology or the like is a set of numerical data, the relationship between the attributes is appropriately set. The effect is that it can be discovered.

  In addition, according to the present invention, when the units of numerical data are adjusted and the units do not match, it is determined that there is no relationship between the sets. Since the statistics related to the distribution of data are calculated, it is possible to appropriately find a case where there is no relationship between sets by checking the unit consistency.

  In addition, according to the present invention, the relationship between sets of numerical data is determined by comparing the distance between sets of numerical data distributions with a predetermined threshold value. There is an effect that it is possible to appropriately find a case where there is no relationship between sets.

  Further, according to the present invention, when there is no overlap between two distributions of the numerical data of the set, the maximum value of the numerical data in the distribution with the smaller value and the minimum value of the numerical data in the distribution with the larger value Since the relationship between sets is determined by checking whether the difference in the number of digits between and is greater than 3 digits, the majority of related sets have a difference in the number of digits of 3. Since it is within digits, there is an effect that it is possible to appropriately find the association between sets.

  In addition, according to the present invention, since the relationship between sets is determined by performing an identity test related to the average value of the numerical data based on the statistic, the equivalence relationship or the equivalence between the sets is determined. There is an effect that it is possible to appropriately discover whether or not there is a relationship.

  Further, according to the present invention, since the significance level related to the test in the case of performing the identity test related to the average value of the numerical data is set within the range of 0.01 to 0.3, By setting the significance level in consideration of the error of the identity test, it is possible to appropriately find the association between sets.

  Further, according to the present invention, since the relevance between sets is determined by performing the identity test related to the dispersion of numerical data based on the statistic, the equivalence relationship between the sets, the equality There is an effect that it is possible to appropriately discover whether there is a relationship or sibling relationship.

  Further, according to the present invention, since the significance level related to the test in the case of performing the identity test related to the dispersion of numerical data is set within the range of 0.01 to 0.3, By setting the significance level in consideration of the error of the sex test, there is an effect that it is possible to appropriately find the association between sets.

Exemplary embodiments of an inter-set relationship determination program and an inter-set relationship determination apparatus according to the present invention will be described below in detail with reference to the accompanying drawings.

  First, the concept of inter-set relationship determination processing according to the present invention will be described. FIG. 1 is a diagram for explaining the concept of inter-set relevance determination processing according to the present invention. FIG. 1 shows a case where the relationship between categories in the two information systems A and B is determined.

  Here, the information system is a classification system composed of categories having an upper-lower hierarchical relationship. The target category here is a concept representing a set having numerical data as elements.

  As shown in FIG. 1, numerical data belonging to a category includes information such as numerical values, multiples, and units. In the example of FIG. 1, the numerical data of the category “red” of the information system A includes a numerical value “X” and a unit “Hz (Hertz)”, and the numerical data of the category “Red” of the information system B is , A numerical value “Y”, a unit magnification “k (kilo)”, and a unit “Hz”.

  In the present embodiment, whether statistical features are extracted from probability distributions of numerical data belonging to categories of different information systems, and is compared, whether there is an equivalence relationship, an equivalence relationship, a sibling relationship, or a hierarchical relationship between the categories. Determine whether or not.

  Here, the equivalence relationship is a relationship in which the categories are the same. An equivalence relationship is a relationship in which categories are not the same but are synonymous. A sibling relationship is a relationship having the same superordinate concept. The hierarchical relationship is a relationship that is a superordinate concept-subordinate concept. Further, the equivalence relation and the equivalence relation are collectively referred to as a similarity relation.

  Here, when extracting a statistical feature from the probability distribution of the numerical data, the statistical feature is extracted after aligning the units of the numerical data. FIG. 2 is a diagram illustrating an example of a probability distribution of numerical data having units.

  In the example of FIG. 2, since the unit “kHz” of the information system B includes a multiple “k”, the unit “Hz” of the information system A and the unit “kHz” of the information system B are different. . For this reason, in this embodiment, the statistical features of the probability distributions of numerical data are compared by aligning units to a predetermined reference unit “Hz”.

  In this way, it becomes possible to automatically detect the relationship between categories of numerical data, etc., so when a company or organization is merged or integrated, the information held by the company or organization is integrated. Can be made easier and work costs can be reduced.

  Here, the information system is a classification system that classifies information into categories. However, an ontology system that defines classes and class attributes having a higher-lower hierarchical relationship, or tables and tables. It may be a database system composed of included fields.

  In the ontology system, a class consisting of a set of numerical data or an attribute consisting of a set of numerical data corresponds to the category, and in the database system, a field consisting of a set of numerical data corresponds to the category.

  FIG. 3 is a diagram illustrating an example in the case of comparing relationships between class attributes. In the case of FIG. 3, the “child” class and the “boy” class each have an age attribute, and numerical data is attached to the age attribute.

  Then, a statistical feature is extracted from the probability distribution of the numerical data and is compared to determine whether there is an equivalence relationship, an equivalence relationship, or a sibling relationship between the “age” attributes.

  In addition, if it is determined that there is an equivalence relationship or an equivalence relationship between the “age” attributes, and a content relationship is found between the other attributes, it is determined that there is an upper and lower hierarchical relationship between the classes to which the attribute belongs. it can.

  FIG. 4 is a diagram for explaining the determination process of the upper and lower hierarchy relationships. As shown in FIG. 4, the “child” class has a gender attribute indicating “male” or “female”, and the “boy” class has a gender attribute “male”. .

  These gender distributions have a containment relationship, but it is difficult to determine the relationship between the classes to which these attributes belong. Therefore, a process of comparing the age attributes of the “child” class and the “boy” class is performed.

  Specifically, a statistical feature is extracted from the probability distribution of numerical data associated with the age attribute, and it is determined whether or not there is an equivalence relationship or an equivalence relationship between the age attributes. When there is an equivalence relationship or an equivalence relationship between age attributes, and there is a containment relationship between other attributes, it is determined that there is an upper and lower hierarchical relationship between classes to which those attributes belong.

  Next, a functional configuration of the inter-set relationship determination apparatus according to the present embodiment will be described. FIG. 5 is a diagram illustrating a functional configuration of the inter-set relationship determination apparatus according to the present embodiment.

  As shown in FIG. 5, the inter-set relationship determination apparatus includes an input unit 10, a display unit 11, a storage unit 12, a unit conversion unit 13, a statistic calculation unit 14, a range interval analysis unit 15, and an average value identity analysis. Section 16, distributed identity analysis section 17, inter-category relation determination section 18, and control section 19.

  The input unit 10 is an input device such as a keyboard or a mouse. The display unit 11 is a display device such as a display. The storage unit 12 is a storage device such as a hard disk device.

  The storage unit 12 stores unit conversion data 12a, first information system data 12b, second information system data 12c, first probability distribution data 12d, and second probability distribution data 12e.

  The unit conversion data 12a is data storing information such as a reference unit for each category of numeric data, other units used as units of numeric data, and unit magnification. The unit conversion data 12a is used to align the unit with the reference unit when the unit used between the category numerical data is different.

  The 1st information system data 12b and the 2nd information system data 12c are data of two different information systems to which the category which judges relevance belongs. The first information system data 12b and the second information system data 12c include hierarchical relationships of categories belonging to the respective information systems, information on numerical data belonging to the categories, and the like.

  The first probability distribution data 12d and the second probability distribution data 12e are statistical data of probability distributions respectively obtained from numerical data belonging to two different information system categories.

  The unit conversion unit 13 performs a process of converting the unit to the reference unit when the unit used between the numerical data of the categories is different.

  Here, the unit is a measure of the numerical data, and it is only after adding a unit to a measurable physical quantity (height, weight, amount of money, speed, milk bottle capacity, etc.) “Real world” is connected (for example, height data 162 cm, amount data 2 million yen, etc.).

  Therefore, the unit is an essential attribute for the numerical data. From an analytical viewpoint, the unit corresponds to the coordinate axis, and the numerical value corresponds to the coordinate value. Then, each real value of the numerical data is distributed on the coordinate axis corresponding to the unit, and it becomes possible to perform a quantitative comparison of each real value.

  Therefore, in order to compare numerical data belonging to two categories, commonality of coordinate axes, that is, commonality of units is indispensable, and numerical data that does not have common coordinate axes cannot be compared.

  However, even if the unit names associated with the numerical data are different, the numerical data can be compared if they are units having the same attribute. For example, there are a plurality of units such as “m (meters)”, “km (kilometers)”, and “cm (centimeters)” as units related to the length, but “k (kilometers)” and “c (centimeters)”. The numerical data can be compared by taking into consideration the difference in the conversion value (unit magnification) between the units such as “)”.

  Therefore, in the present embodiment, the unit conversion unit 13 converts the unit of the numerical category into a reference unit. Here, the reference unit is predetermined as “m” in the case of the “length” attribute, “g” in the case of the “weight” attribute, or the like.

  Specifically, the unit conversion unit 13 reads the first information system data 12b and the second information system data 12c from the storage unit 12, and the category name, the unit name associated with the numerical data in the category, the unit multiple, Get real value information of all numeric data belonging to a category.

  Then, for example, when the units of the two numerical categories are “m” and “cm”, the unit conversion unit 13 uses the unit of “m” as it is because it is a reference unit. The unit “c” is converted to a reference unit “m” by applying a unit magnification 0.01 of “c” to numerical data having a unit of “cm”.

If unit name U, unit magnification M, real value x, and reference unit name U ^* , the above processing is
x | U ⇒ x ・ M | U ^*
And can be expressed in general.

Here, the category name of two categories "A" and "B", the unit name U _A and U _B, real values x _1, ···, x _m and y _1, · · ·, and y _n When represent respectively, the real value of the scaled two categories were _{the, x 1 · M a, ···} , x m · M a and _{y 1 · M B, ···,} y n · M B next The unit names are U _A ^* and U _B ^* .

The statistic calculation unit 14 performs a process of calculating the statistic of the probability distribution related to the numerical data of the two categories. Specifically, using the following formulas (1) to (6), the number of data n _A and n _{B of} the probability distribution relating to the numerical data of the two categories “A” and “B”, the average value μ _A and Statistics such as μ _B , variances σ _A ² and σ _B ² are calculated.

Here, as the data numbers n _A and n _B of the numerical data, the known data numbers m and n are set as they are. Then, the statistic calculator 14 stores the statistics of the two probability distributions in the storage unit 12 as the first probability distribution data 12d and the second probability distribution data 12e.

  The value range analysis unit 15 performs processing for analyzing the difference between the value ranges of the two probability distributions. For example, if the categories “A” and “B” are categories having an equivalence relationship or an equivalence relationship, the range of probability distributions is similar to the probability distribution of the categories “red” and “Red” shown in FIG. Should overlap quite a bit.

  Therefore, when there is an equivalence relationship or an equivalence relationship between categories, it is a necessary condition that the range of probability distributions overlap.

  In addition, when there is a sibling relationship between two categories, it is conceivable that the two probability distribution ranges are close to each other or overlap (although slightly). Therefore, in order for the two probability distributions to have a sibling relationship, it is necessary that the respective value ranges overlap (although slightly) or be distributed close to each other.

  FIG. 6 is a diagram illustrating an example of a probability distribution of categories in a sibling relationship. The example of FIG. 6 shows a case where the frequency category “red” and the frequency category “blue” are in a sibling relationship with the common concept “frequency” as a parent concept. As described above, the value ranges of the probability distributions of the frequency category “red” and the frequency category “blue” are distributed close to each other.

In order to investigate such overlapping of the range, the range separation analysis unit 15 performs the upper limit value (maximum value) x ₊ and lower limit value (minimum value) x ₋ of the category A probability distribution X (A), and the probability of category B. Information on the upper limit value y ₊ and the lower limit value y ₋ of the distribution X (B) is acquired, and it is checked whether or not the following conditions are satisfied.

(X ₋ ≦ y ₋ <x ₊ ) ∨ (x ₋ <y ₊ ≦ x ₊ )

  In addition, the value range analysis unit 15 determines whether or not the value ranges of the two probability distributions are close according to the following conditional expressions (7) to (18). In the following expression, X is a range of category “A”, and Y is a range of category “B”.

Here, O ₁ is a threshold value for determining whether or not the value range is close. As shown in the equations (7) to (18), if the ratio of the upper limit value and the lower limit value of the log scale is smaller than O ₁ , the two probability distribution ranges are determined to be close and correspond to the respective probability distributions. The category is determined to have a sibling relationship.

Conversely, if the ratio of the upper limit value and the lower limit value of the log scale is greater than O ₁ , the two probability distribution ranges are determined to be far, and the category corresponding to each probability distribution has an equivalence relationship, an equality relationship, or It is determined that there is no sibling relationship.

The value of O ₁ is set to 3 or less. This is because the numerical data of the category having the majority of sibling relationships falls within the difference in the number of digits of about three digits.

  The average value identity analysis unit 16 performs processing to test whether or not the average values of the two probability distributions have identity. Specifically, the average value identity analysis unit 16 performs an average value identity test by Welch's test. Welch's test is a test method used when equal distribution of probability distribution cannot be assumed in advance.

First, the average value identity analysis unit 16 sets a hypothesis. That is, the average value identity analysis unit 16 sets the significance level to a%, the null hypothesis H0 to μ _A = μ _B , and the alternative hypothesis H1 to μ _A ≠ μ _B. Then, the average value identity analysis unit 16 calculates the t value t ₀ and the degree of freedom ν by the following equations (19) and (20).

Then, the average value identity analysis unit 16 calculates _a value t _{1−a / 2} (ν) at both sides a% point of the t distribution function. FIG. 7 is a diagram for explaining the value t _{1−a / 2} (ν) of the a% point on both sides of the t distribution function. As shown in FIG. 7, here, since the significance level is a%, the value of the upper a / 2% point, that is, t ^* _a (ν) = t _{1−a / 2} (ν) is calculated. .

Thereafter, the average value identity analysis unit 16 examines the magnitude relationship between the absolute value | t ₀ | of the t value and t _{1−a / 2} (ν), and determines whether or not there is a significant difference in the null hypothesis H0. Find out.

Specifically, the average value identity analysis unit 16 determines that there is a significant difference when | t ₀ | ≧ t _{1−a / 2} (ν), and rejects the null hypothesis H0. The alternative hypothesis H1 is adopted. If | t ₀ | <t _{1−a / 2} (ν), the mean value identity analysis unit 16 determines that there is no significant difference, rejects the alternative hypothesis H1 by adopting the null hypothesis H0. To do.

  In this way, if the null hypothesis H0 is adopted, the identity of the average values of the two probability distributions is not denied, and the equivalence between the two categories “A” and “B” corresponding to the respective probability distributions. There is a possibility that there is a relationship or an equivalence relationship.

  If the alternative hypothesis H1 is adopted, the identity of the average values of the two probability distributions is denied, and there is an equivalence relationship or an equivalence relationship between the two categories “A” and “B” corresponding to each probability distribution. The possibility is also denied.

  The variance identity analysis unit 17 performs processing to test whether or not the variances of the two probability distributions are identical. Specifically, the variance identity analysis unit 17 tests the identity of variance by F test.

First, the variance identity analysis unit 17 sets a hypothesis. That is, the variance identity analysis unit 17 sets the significance level to b%, the null hypothesis H0 to σ _A ² = σ _B ² , and the alternative hypothesis H1 to σ _A ² ≠ σ _B ² .

Then, the variance identity analysis unit 17 reads the information of the variances σ _A ² and σ _B ² calculated by the statistic calculation unit 14 from the first probability distribution data and the second probability distribution data stored in the storage unit 12. The larger one is set to V ₁ and the smaller one is set to V ₂ .

In addition, the variance identity analysis unit 17 reads information on the data numbers n _A and n _B from the first probability distribution data and the second probability distribution data, and sets them as the sample numbers n ₁ and n ₂ . Then, the variance identity analysis unit 17 calculates the degrees of freedom ν ₁ and ν ₂ and the variance ratio F ₀ (> 1) by the following equations (21) and (22).

Then, the variance identity analysis unit 17 calculates the value F _{1−b / 2} (ν ₁ , ν ₂ ) of the upper probability b / 2% point of the F distribution function. FIG. 8 is a diagram for explaining the value F _{1−b / 2} (ν ₁ , ν ₂ ) of the upper probability b / 2% point of the F distribution function. As shown in FIG. 8, here, since the significance level is b%, the value of the upper b / 2% point, that is, F ^* _b (ν ₁ , ν ₂ ) = F _{1−b / 2} (ν ₁ , ν ₂ ).

Thereafter, the variance identity analysis unit 17 examines the magnitude relationship between the variance ratio F ₀ and F _{1−b / 2} (ν ₁ , ν ₂ ) to determine whether there is a significant difference in the null hypothesis H0. Investigate.

Specifically, the variance identity analysis unit 17 determines that there is a significant difference when F ₀ ≧ F _{1−b / 2} (ν ₁ , ν ₂ ), and rejects the null hypothesis H0. The alternative hypothesis H1 is adopted. When F ₀ <F _{1−b / 2} (ν ₁ , ν ₂ ), the variance identity analysis unit 17 determines that there is no significant difference, adopts the null hypothesis H0, and sets the alternative hypothesis H1. Dismiss.

  Thus, if the null hypothesis H0 is adopted, the identity of the variance of the two probability distributions cannot be denied. On the other hand, if the alternative hypothesis H1 is adopted, the identity of the variance of the two probability distributions is denied.

  The inter-category relevance determination unit 18 is based on the results analyzed by the unit conversion unit 13, the range separation analysis unit 15, the average value identity analysis unit 16, and the variance identity analysis unit 17, between the categories of two numerical data. Is processed to determine whether or not there is an equivalence relation, an equality relation, a sibling relation, or an upper and lower hierarchy relation.

Specifically, the inter-category relevance determination unit 18, when the unit of numerical data is converted into the reference unit by the unit conversion unit 13, the reference unit names U _A ^{* of the} two categories “A” and “B” ^. and it is determined whether U _B ^* is equal, if ^* reference unit name U _a ^* and U _B are not equal, the relationship between the two categories is determined that there is no, and outputs the result.

  In addition, the inter-category relevance determination unit 18 determines that the reference unit names of the two categories match, the two probability distribution ranges of the numerical data overlap, and the average value and variance are the same. In this case, it is determined that there is an equivalence relationship between the two categories, and the result is output.

  In addition, the inter-category relevance determination unit 18 matches the reference unit names of the two categories, the two probability distribution ranges of the numerical data overlap, and the average value is the same although the identity of the variance is denied. If it is determined that there is, it is determined that there is an equivalence relationship between the two categories, and the result is output.

  In addition, the inter-category relevance determination unit 18 matches the reference unit names of the two categories, the two probability distribution ranges of the numerical data are overlapped, and the mean value identity is denied but the variance is the same. If it is determined that there is a sibling relationship between the two categories, the result is output.

  Further, the inter-category relevance determination unit 18 matches the reference unit names of the two categories and does not overlap the ranges of the two probability distributions of the numerical data, but is close to the range of values according to the equations (7) to (18). If it is determined that the distribution is the same, it is determined that there is a sibling relationship between the two categories, and the result is output.

  In other cases, the inter-category relationship determination unit 18 determines that there is no relationship between the two categories, and outputs the result.

  The control unit 19 is a control unit that controls the inter-group relevance determination device as a whole, and controls data exchange between the functional units.

  Next, a processing procedure of inter-set relevance determination processing according to the present embodiment will be described. FIGS. 9A and 9B are flowcharts (1) and (2) for explaining the processing procedure of the inter-set relevance determination processing according to the present embodiment.

  As shown in FIG. 9A, first, the unit conversion unit 13 of the inter-group relevance determination device reads unit conversion data 12a, first information system data 12b, and second information system data 12c from the storage unit 12 (steps). S101).

  And the unit conversion part 13 adjusts the unit of the numerical data of the category which belongs to two information systems, and determines whether it is relevant or not to a reference unit (step S102). Thereafter, the inter-category relationship determination unit 18 checks whether or not the adjusted units match (step S103).

  When the units do not match (No at Step S103), as shown in FIG. 9B, the inter-category relationship determination unit 18 determines that the two categories are irrelevant (Step S117), and displays the determination result. This is output to the unit 11 (step S118), and the inter-set relevance determination process is terminated.

  When the units match (step S103, Yes), as shown in FIG. 9A, the mean value identity analysis unit 16 and the variance identity analysis unit 17 set the significance levels a and b, and Then, a quantile corresponding to the significance level is obtained to calculate a significant value range (step S104).

  Thereafter, the statistic calculation unit 14 calculates the number of data, the average value, and the variance relating to the numerical data of the two probability distributions using the formulas (1) to (6), and stores the first probability distribution data in the storage unit 12. 12d and the second probability distribution data 12e are stored (step S105).

Then, as shown in FIG. 9-2, the value range analysis unit 15 determines whether the value ranges of the probability distributions of the two types of numerical data overlap, the absolute value of the probability distribution interval is O ₁ or less, or Then, it is examined whether the absolute value of the probability distribution interval is larger than O ₁ (step S106).

  If the value ranges of the probability distributions of the two types of numerical data overlap (step S106, “overlap”), the average value identity analysis unit 16 performs an identity test of the average values of the two types of numerical data. It is executed (step S107), and it is checked whether the average values are the same (step S108).

  If the average values are the same (step S108, Yes), the variance identity analysis unit 17 performs an identity test of the variance of the two types of numerical data (step S109), and whether the variances are the same. (Step S110).

  If the variances are the same (step S109, Yes), the inter-category relationship determination unit 18 determines that the two categories are the same concept (step S111), and outputs the determination result to the display unit 11. (Step S113), this inter-set relevance determination process is terminated.

  When the variances are not the same in step S110 (No in step S109), the inter-category relationship determination unit 18 determines that the two categories are equal concepts (step S112), and the determination result is displayed on the display unit 11. This is output (step S113), and the inter-set relevance determination process is terminated.

If the average values are not the same in step S108 (step S108, No), or if the absolute value of the difference between the probability distributions of the two types of numerical data is less than O _{1 in} step S106 (step S106, “| Distance | ≦ O ₁ )), the variance identity analysis unit 17 executes a variance identity test of two types of numerical data (step S114), and checks whether the variances are the same (step S115).

  When the variances are the same (step S115, Yes), the inter-category relationship determination unit 18 determines that the two categories are sibling concepts (step S116), and outputs the determination result to the display unit 11. (Step S113), this inter-set relevance determination process is terminated.

If the variances are not the same in step S115 (step S115, No), or if the absolute value of the difference between the probability distributions related to the two types of numerical data is greater than O _{1 in} step S106 (step S106, “| > O ₁ ”), the variance identity analysis unit 17 determines that the two categories are irrelevant (step S 117), outputs the determination result to the display unit 11 (step S 113), and determines the association between the sets. The process ends.

  Here, it is determined whether there is an equivalence relationship, an equivalence relationship, or a sibling relationship between the categories. However, as described with reference to FIG. It may be determined whether there is a hierarchical relationship between classes based on the relationship between them.

  In this case, the inter-category relationship determination unit 18 checks whether there is an equivalence relationship or an equivalence relationship between attributes according to the method described above. Further, the inter-category relevance determination unit 18 checks whether or not other attributes have a content relationship.

  Then, the inter-category relevance determination unit 18 has an upper or lower hierarchy between classes to which these attributes belong when there is an equivalence relationship or an equivalence relationship between attributes consisting of numerical data and there is a content relationship between other attributes. Determine that there is a relationship.

  Further, when an equivalence relationship is found between other attributes, the inter-category relationship determination unit 18 determines whether the relationship between attributes consisting of numerical data is an equivalence relationship or an equivalence relationship. It is determined that there is an equivalence relationship or an equivalence relationship between classes.

  Further, the inter-category relevancy determination unit 18 determines that there is an equivalence relationship between other attributes, and the relationship between attributes composed of numerical data is an equivalence relationship or an equivalence relationship. It is determined that there is an equivalence relationship.

  Incidentally, the various processes described in the above embodiments can be realized by executing a program prepared in advance by a computer. Therefore, in the following, an example of a computer that executes a program that realizes the various processes will be described with reference to FIG.

  FIG. 10 is a diagram illustrating a hardware configuration of a computer serving as the inter-group relationship determination apparatus illustrated in FIG.

  The computer includes an input device 100 that receives input of data from a user, a display device 101, a RAM (Random Access Memory) 102, a ROM (Read Only Memory) 103, and a medium reading device that reads a program from a recording medium in which various programs are recorded. 104, a network interface 105 that exchanges data with other computers via a network, a CPU (Central Processing Unit) 106, and an HDD (Hard Disk Drive) 107 are connected by a bus 108.

  The HDD 108 stores a program that exhibits the same function as that of the inter-group relevance determination apparatus, that is, an inter-group relevance determination program 107b. Here, the inter-set relationship determination program 107b may be appropriately distributed and stored.

  Then, the CPU 106 reads out and executes the inter-group relationship determination program 107b from the HDD 107, thereby functioning as the inter-group relationship determination process 106a.

  This inter-set relationship determination process 106a includes a unit conversion unit 13, a statistic calculation unit 14, a range interval analysis unit 15, an average value identity analysis unit 16, a variance identity analysis unit 17, and an inter-category relationship shown in FIG. This corresponds to the sex determination unit 18 and the control unit 19.

  The HDD 107 stores various data 107a. The various data 107a includes unit conversion data 12a, first information system data 12b, second information system data 12c, first probability distribution data 12d, and second probability distribution data 12e stored in the storage unit 12 shown in FIG. Corresponding to

  The CPU 106 stores various data 107 a in the HDD 107, reads the various data 107 a from the HDD 107, stores it in the RAM 102, and executes data processing based on the various data 102 a stored in the RAM 102.

  By the way, the inter-set relationship determination program 107b does not necessarily need to be stored in the HDD 107 from the beginning.

  For example, a “portable physical medium” such as a flexible disk (FD), a CD-ROM, an MO disk, a DVD disk, a magneto-optical disk, and an IC card inserted into a computer, or a hard disk drive (inside and outside the computer) Each program is stored in a “fixed physical medium” such as an HDD), and “another computer (or server)” connected to the computer via a public line, the Internet, a LAN, a WAN, or the like. The computer may read and execute each program from these.

  Although the embodiments of the present invention have been described so far, the present invention may be implemented in various different embodiments in addition to the above-described embodiments within the scope of the technical idea described in the claims. It ’s good.

  For example, in the above embodiment, the inter-set relationship determination device determines whether there is an equivalence relationship, an equivalence relationship, a sibling relationship, or an upper / lower hierarchy relationship between categories. May only determine whether one of those relationships is between categories.

  In addition, among the processes described in this embodiment, all or part of the processes described as being performed automatically can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method.

  In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

  Each component of the illustrated inter-set relationship determination apparatus is functionally conceptual and does not necessarily need to be physically configured as illustrated.

  In other words, the specific form of distribution / integration of the inter-set relationship determination device is not limited to the one shown in the figure, and all or a part thereof can be functionally or physically processed in an arbitrary unit according to various loads or usage conditions. Can be distributed and integrated.

  Further, each processing function performed in the inter-set relationship determination apparatus is realized by a CPU and a program that is analyzed and executed by the CPU, or as hardware by wired logic. Can be realized.

(Supplementary note 1) A program for determining relevance between sets for determining relevance between sets having numerical data as elements,
A statistic calculation procedure for calculating a statistic related to the distribution of the numerical data;
A statistic storage procedure for storing a statistic calculated by the statistic calculation procedure;
A relevance determination procedure for determining relevance between sets based on the statistics stored in the statistics storage procedure;
A program for determining the relationship between sets, which causes a computer to execute.

(Additional remark 2) The said relevance determination procedure determines whether any one of equivalence relations, equivalence relations, sibling relations, or upper and lower hierarchy relations exists between sets. Program for determining association between sets described in 1.

(Additional remark 3) The said set is a set which concerns on the attribute of a predetermined concept, The relevance determination program between sets as described in Additional remark 1 or 2 characterized by the above-mentioned.

(Supplementary Note 4) If the units of the numerical data are adjusted and the units do not match, the computer further executes a unit adjustment procedure for determining that there is no relationship between the sets, and the statistical calculation procedure includes: 4. The inter-group relevance determination program according to appendix 1, 2, or 3, wherein, when the units match, a statistic relating to a distribution of numerical data whose units are adjusted by the unit adjustment procedure is calculated. .

(Additional remark 5) The said relevance determination procedure determines the relevance between sets by comparing the gap between distribution of the numerical data of a set with a predetermined threshold value, Any one of Additional remarks 1-4 characterized by the above-mentioned. The program for determining the relationship between sets described in 1.

(Additional remark 6) The said relevance determination procedure is the case where there is no overlap between the two distributions of the numerical data of the set, the maximum value of the numerical data in the distribution with the smaller value, and the numerical data in the distribution with the larger value 6. The inter-set relevance determination program according to appendix 5, wherein the relevance between sets is determined by examining whether or not the difference in the number of digits from the minimum value of the set is greater than 3 digits.

(Supplementary Note 7) The relevance determination procedure determines relevance between sets by performing an identity test on the average value of the numerical data based on the statistic stored in the statistic storage procedure. The inter-group relevance determination program according to any one of Supplementary notes 1 to 6, characterized in that:

(Supplementary Note 8) The relevance determination procedure is characterized in that a significance level related to a test in the case of performing an identity test related to an average value of the numerical data is set within a range of 0.01 to 0.3. The program for determining relevance between sets according to appendix 7.

(Supplementary Note 9) In the relevance determination procedure, the relevance between sets is determined by performing an identity test related to the dispersion of the numerical data based on the statistic stored in the statistic storage procedure. The inter-group relevance determination program according to any one of Supplementary notes 1 to 8, characterized in that:

(Additional remark 10) The said relevance determination procedure sets the significance level which concerns on the test in the case of performing the test of the identity which concerns on dispersion | distribution of the said numerical data within the range of 0.01 to 0.3, It is characterized by the above-mentioned. The program for determining relevance between sets according to appendix 9.

(Supplementary note 11) A method for determining the relationship between sets for determining the relationship between sets having numerical data as an element,
A statistic calculating step of calculating a statistic related to the distribution of the numerical data;
A statistic storage step for storing the statistic calculated by the statistic calculation step;
A relevance determination step of determining relevance between sets based on the statistic stored in the statistic storage step;
A method for determining the relationship between sets, characterized in that

As described above, the inter-set relevance determination program and inter-set relevance determination apparatus according to the present invention are required to appropriately find the relevance between sets having numerical data as elements. Useful for.

It is a figure explaining the concept of the relationship determination process between sets which concerns on this invention. It is a figure which shows an example of probability distribution of the numerical data which has a unit. It is a figure which shows an example in the case of comparing the relationship between the attributes of a class. It is a figure explaining the determination process of a high-order hierarchy relation. It is a figure explaining the functional structure of the relationship determination apparatus between sets which concerns on a present Example. It is a figure which shows an example of the probability distribution of the category which has a sibling relationship. It is a figure explaining the value t1 _{-a / 2} ((nu)) of the both sides a% point of t distribution function. It is a figure explaining value F1 _{-b / 2} ((nu) ₁ , (nu) ₂ ) of the upper probability b / 2% point of F distribution function. It is a flowchart (1) explaining the process sequence of the relationship determination process between sets which concerns on a present Example. It is a flowchart (2) explaining the process sequence of the relationship determination process between sets which concerns on a present Example. It is a figure which shows the hardware constitutions of the computer used as the inter-group relationship determination apparatus shown in FIG. It is a figure explaining the difference of the category between different classification systems.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Input part 11 Display part 12 Storage part 12a Unit conversion data 12b 1st information system data 12c 2nd information system data 12d 1st probability distribution data 12e 2nd probability distribution data 13 Unit conversion part 14 Statistics calculation part 15 Value range analysis Unit 16 Average value identity analysis unit 17 Variance identity analysis unit 18 Inter-category relationship determination unit 19 Control unit

Claims (5)

A program for determining relevance between sets for determining relevance between sets having numerical data as an element,
A statistic calculation procedure for calculating a statistic related to the distribution of the numerical data;
A statistic storage procedure for storing a statistic calculated by the statistic calculation procedure;
The relationship between sets is determined based on the statistics stored in the statistics storage procedure, and the ranges of the probability distributions of the numerical data between the sets overlap, but are identical to the average value of the probability distribution of the numerical data between the sets If the absolute value of the range of the probability distribution of numerical data between sets is less than or equal to a predetermined threshold and the variance of the probability distribution of numerical data between sets is identical, the sibling concept Relevance determination procedure for determining that there is,
A program for determining the relationship between sets, which causes a computer to execute.   If the units of the numerical data are adjusted and the units do not match, the computer further executes a unit adjustment procedure for determining that there is no relationship between the sets, and the statistics calculation procedure matches the units. 2. The inter-set relevance determination program according to claim 1, wherein a statistic relating to a distribution of numerical data whose units are adjusted by the unit adjustment procedure is calculated.   The relevance determination procedure is characterized in that the relevance between sets is determined by performing an identity test on an average value of the numerical data based on the statistic stored in the statistic storage procedure. The inter-group relevance determination program according to claim 1 or 2.   The relevance determination procedure is characterized in that the relevance between sets is determined by performing an identity test related to the dispersion of the numerical data based on the statistics stored in the statistical storage procedure. The inter-set relationship determination program according to any one of claims 1 to 3. A device for determining the relationship between sets that determines the relationship between sets having numerical data as an element,
Statistic calculation means for calculating a statistic related to the distribution of the numerical data;
Statistic storage means for storing the statistic calculated by the statistic calculation means;
The relationship between sets is determined based on the statistics stored in the statistic storage means, and the ranges of the probability distributions of the numerical data between the sets overlap, but they are identical to the average value of the probability distribution of the numerical data between the sets If the absolute value of the range of the probability distribution of numerical data between sets is less than or equal to a predetermined threshold and the variance of the probability distribution of numerical data between sets is identical, the sibling concept Relevance determining means for determining that there is,
A device for determining the relationship between sets, comprising:

Images