JP2012181605A - Data analysis support device and program - Google Patents

Abstract

An object of the present invention is to provide a data analysis support apparatus and program capable of associating appropriate attributes between arbitrary data tables.
A first word extracting unit extracts a first word from an attribute value of a first character string type attribute constituting a first data table. The second word extracting means extracts the second word from the attribute value of the second character string type attribute constituting the second data table. The similarity calculation means is configured to output a first character string constituting the first data table based on the first word extracted by the first extraction means and the second word extracted by the second extraction means. The similarity between the type attribute and the second character string type attribute constituting the second data table is calculated. The similar attribute candidate extraction unit extracts the first character string type attribute and the second character string type attribute as similar attribute candidates based on the similarity calculated by the similarity calculation unit.
[Selection] Figure 2

Description

  Embodiments described herein relate generally to a data analysis support apparatus and program for associating attributes representing the same content among a plurality of data tables.

  For example, in order to analyze business mistakes at a plurality of banks or the like, refer to different data tables for each bank in which data (information) related to the business mistakes is stored, and set attributes that represent the same contents between the data tables. It is necessary to compare.

  However, these data tables are generally defined differently in each bank. For example, the attribute names or attribute values are often expressed differently.

  Therefore, it is difficult to associate attributes representing the same contents between a plurality of data tables having different definitions.

  In this connection, for example, by comparing rows having the same primary key value between two data tables having the same primary key in the data table, and calculating the matching rate of attributes other than the primary key, A technique for extracting attributes common to both data tables (hereinafter referred to as a first technique) is known.

  Also, for example, the similarity of attribute name, attribute value distribution, attribute value grapheme distribution and attribute value string length distribution (that is, attribute features) is calculated, and corresponding attributes are extracted between data tables. Technology (hereinafter referred to as second technology) is known.

Japanese Patent Laid-Open No. 2004-86782 JP 2003-271656 A

  However, according to the first technique described above, for example, the primary keys of two data tables must be common. Therefore, in the first technique, similar attributes cannot be extracted between arbitrary data tables unless the primary keys are common.

  On the other hand, according to the second technique described above, since only the feature of the attribute to be compared is used, it is possible to extract a similar attribute between arbitrary data tables. However, according to the second technique, for example, in the case of a character string type attribute (that is, an attribute having an attribute value including a character string), the meaning of the attribute value is not taken into consideration. There is a case that cannot be attached.

  Therefore, the problem to be solved by the present invention is to provide a data analysis support apparatus and program capable of associating appropriate attributes between arbitrary data tables.

  The data analysis support apparatus according to the embodiment includes a data table storage unit, a first word extraction unit, a second word extraction unit, a similarity calculation unit, and a similar attribute candidate extraction unit.

  The data table storage means includes a first data table including a first attribute including a first character string type attribute having an attribute value including a character string, and a second character string including an attribute value including the character string. A second data table composed of second attributes including type attributes is stored in advance.

  The first word extraction unit is configured to extract a character string included in the attribute value of the first character string type attribute included in the first attribute included in the first data table stored in the data table storage unit. A first word to be configured is extracted.

  A second word extracting unit that extracts a character string included in an attribute value of a second character string type attribute included in a second attribute included in the second data table stored in the data table storing unit; A second word to be configured is extracted.

  The similarity calculation means includes a first data table constituting the first data table based on the first word extracted by the first extraction means and the second word extracted by the second extraction means. The similarity between the first character string type attribute included in the second attribute and the second character string type attribute included in the second attribute constituting the second data table is calculated.

  The similar attribute candidate extracting unit configures the first character string type attribute included in the first attribute configuring the first data table and the second data table based on the calculated similarity. A second character string type attribute included in the second attribute is extracted as a similar attribute candidate.

The block diagram which shows the hardware constitutions of the data analysis assistance apparatus which concerns on 1st Embodiment. The block diagram which mainly shows a function structure of the data analysis assistance apparatus 30 shown in FIG. The block diagram which shows the function structure of the numerical type attribute process part 32 shown in FIG. The block diagram which shows the function structure of the character string type | mold attribute process part 33 shown in FIG. The figure which shows an example of the data structure of the data table of A bank stored in the data table storage part 22. FIG. The figure which shows an example of the data structure of the data table of B bank stored in the data table storage part 22. FIG. The figure which shows an example of the data structure of the data table of C bank stored in the data table storage part 22. FIG. The flowchart which shows the process sequence of the data analysis assistance apparatus 30 which concerns on this embodiment. The flowchart which shows the process sequence of the similarity calculation process of the numerical attribute contained in a similarity calculation process. The figure for demonstrating concretely the similarity between the 1st and 2nd numerical type attributes calculated by the numerical type attribute similarity calculation part 323. FIG. The figure for demonstrating concretely the similarity between the 1st and 2nd numerical type attributes calculated by the numerical type attribute similarity calculation part 323. FIG. The figure which shows an example of the similarity list created in the similarity calculation process of a numerical type attribute. The flowchart which shows the process sequence of the similarity calculation process of the character string type attribute contained in a similarity calculation process. The figure for demonstrating concretely the word set of the target character string type | mold attribute produced by the attribute word extraction part 331. FIG. The figure for demonstrating concretely the similarity between the 1st and 2nd character string type | mold attribute calculated by the character string type | mold attribute similarity calculation part 333. FIG. The figure which shows an example of the similarity list created in the similarity calculation process of a character string type | mold attribute. The flowchart which shows the process sequence of a similar attribute candidate extraction process. The figure for demonstrating concretely about the similarity between the attribute name of the target attribute calculated by the similar attribute candidate extraction part 35, and the attribute name of an applicable attribute. The figure for demonstrating concretely about the similarity between the attribute name of the target attribute calculated by the similar attribute candidate extraction part 35, and the attribute name of an applicable attribute. The figure which shows an example of the data structure of the similar attribute candidate storage part. The block diagram which shows the function structure of the character string type | mold attribute process part 33 contained in the data analysis assistance apparatus 30 which concerns on 2nd Embodiment. The figure for demonstrating concretely the similarity between the 1st and 2nd character string type | mold attribute calculated by the character string type | mold attribute similarity calculation part 335. FIG.

  Hereinafter, each embodiment will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a block diagram illustrating a hardware configuration of the data analysis support apparatus according to the first embodiment. As shown in FIG. 1, the computer 10 is connected to an external storage device 20 such as a hard disk drive (HDD). The external storage device 20 stores a program 21 executed by the computer 10. The computer 10 and the external storage device 20 constitute a data analysis support device 30.

  This data analysis support device 30 is used, for example, for associating attributes among a plurality of different data tables (data tables having different definitions) when analyzing data.

  FIG. 2 is a block diagram mainly showing a functional configuration of the data analysis support apparatus 30 shown in FIG.

  As shown in FIG. 2, the data analysis support device 30 includes an attribute type classification unit 31, a numeric type attribute processing unit 32, a character string type attribute processing unit 33, a threshold value input unit 34, and a similar attribute candidate extraction unit 35. In the present embodiment, these units 31 to 35 are realized by the computer 10 illustrated in FIG. 1 executing the program 21 stored in the external storage device 20. This program 21 can be stored in advance in a computer-readable storage medium and distributed. Further, this program 21 may be downloaded to the computer 10 via, for example, a network.

  In addition, the data analysis support device 30 includes a data table storage unit 22, a numerical attribute storage unit 23, a character string attribute storage unit 24, a numerical attribute similarity storage unit 25, a character string attribute similarity storage unit 26, and a similarity. An attribute candidate storage unit 27 is included. In the present embodiment, these units 22 to 27 are stored in, for example, the external storage device 20.

  The data table storage unit 22 stores a plurality of different data tables (first and second tables) to be analyzed. Each of the plurality of data tables stored in the data table storage unit 22 includes attributes. Further, the attributes constituting the data table include, for example, a numeric attribute and a character string attribute. The numeric type attribute is an attribute having an attribute value including a numeric value (that is, a numeric value can be taken as the attribute value). On the other hand, the character string type attribute is an attribute having an attribute value including a character string (that is, a character string can be taken as an attribute value). Note that the character string included in the attribute value of the character string type attribute is composed of, for example, a word.

  The attribute type classification unit 31 classifies the attributes constituting each of the data tables stored in the data table storage unit 22 into numerical value type attributes or character string type attributes.

  In the numeric attribute storage unit 23 and the character string attribute storage unit 24, the classification results by the attribute type classification unit 31 are stored. Specifically, the numeric type attribute storage unit 23 has, for each data table stored in the data table storage unit 22, a numeric type attribute (attribute name and attribute value) among the attributes constituting the data table. Stored. The character string type attribute storage unit 24 stores, for each data table stored in the data table storage unit 22, a character string type attribute (attribute name and attribute value) among the attributes constituting the data table. Is done.

  The numerical value attribute processing unit 32 is an attribute value (a numerical value included) included in two numerical value attributes (first and second numerical value attributes) constituting different data tables stored in the numerical value attribute storage unit 23. Based on the above, the similarity between the two numeric attributes is calculated. The numeric attribute processing unit 32 calculates the similarity for all combinations of two numeric attributes constituting different data tables stored in the numeric attribute storage unit 23.

  The numeric attribute similarity storage unit 25 stores the similarity for each combination of two numeric attributes constituting different data tables calculated by the numeric attribute processor 32.

  The character string type attribute processing unit 33 includes attribute values (first and second character string type attributes) included in different data tables stored in the character string type attribute storage unit 24. The similarity between the two character string type attributes is calculated based on the included character string. The character string type attribute processing unit 33 calculates the similarity for all combinations of two character string type attributes that constitute different data tables stored in the character string type attribute storage unit 24.

  The character string type attribute similarity storage unit 26 stores the similarity for each combination of two character string type attributes constituting different data tables calculated by the character string type attribute processing unit 33.

  The threshold value input unit 34 inputs a threshold value used when extracting attribute candidates (hereinafter referred to as similar attribute candidates) representing the same contents between different data tables. The threshold value input by the threshold value input unit 34 is specified by the user, for example.

  The similar attribute candidate extraction unit 35 is based on the similarity for each combination of two numeric type attributes constituting different data tables stored in the numeric type attribute similarity storage unit 25 and the threshold value input by the threshold value input unit 34. The two numeric type attributes are extracted as similar attribute candidates. Further, the similar attribute candidate extraction unit 35 is input by the similarity and threshold value input unit 34 for each combination of two character string type attributes constituting different data tables stored in the character string type attribute similarity storage unit 26. Based on the threshold, the two character string type attributes are extracted as similar attribute candidates.

  The similar attribute candidate storage unit 27 stores similar attribute candidates extracted by the similar attribute candidate extraction unit 35.

  FIG. 3 is a block diagram showing a functional configuration of the numerical attribute processing unit 32 shown in FIG. As shown in FIG. 3, the numeric attribute processing unit 32 includes an attribute value numeric range specifying unit 321, an attribute value numeric range storage unit 322, and a numeric attribute similarity calculating unit 323.

  The attribute value numerical value range specifying unit 321 specifies a numerical value range (hereinafter referred to as a numerical value range of the attribute value) included in the attribute value of the numerical value attribute stored in the numerical value attribute storage unit 23. The attribute value numerical value range specifying unit 321 specifies the numerical value range of the attribute value for all the numerical value attributes stored in the numerical value attribute storage unit 23.

  The attribute value numerical value range storage unit 322 stores the numerical value range of the attribute value that the numerical value attribute specified by the attribute value numerical value range specification unit 321 has.

  The numerical value type attribute similarity calculation unit 323 is configured to calculate the similarity between two numerical type attributes constituting different data tables based on the numerical value range of the attribute value of each numerical value attribute stored in the attribute value numerical value range storage unit 322. Calculate the degree. Specifically, the numerical attribute similarity calculation unit 323 calculates the similarity between the two numerical attributes based on a range in which the numerical ranges of the attribute values of the two numerical attributes constituting the different data tables overlap. Is calculated. Note that the numerical value attribute similarity calculation unit 323 calculates similarity for all combinations of two numerical value attributes that constitute different data tables as described above. The similarity calculated by the numerical attribute similarity calculator 323 is stored in the numerical attribute similarity storage 25.

  FIG. 4 is a block diagram showing a functional configuration of the character string type attribute processing unit 33 shown in FIG. As shown in FIG. 4, the character string type attribute processing unit 33 includes an attribute value word extraction unit 331, an attribute value word set storage unit 332, and a character string type attribute similarity calculation unit 333.

  The attribute value word extraction unit 331 performs morphological analysis on the character string included in the attribute value of the character string type attribute stored in the character string type attribute storage unit 24. The attribute value word extraction unit 331 extracts words constituting the character string included in the attribute value of the character string type attribute stored in the character string type attribute storage unit 24 based on the morphological analysis result. Thereby, the attribute value word extraction unit 331 creates a set of words constituting the character string included in the attribute value of the character string type attribute (hereinafter simply referred to as a word set of the attribute value). Note that the attribute value word extraction unit 331 creates a word set of attribute values for all the character string type attributes stored in the character string type attribute storage unit 24.

  The attribute value word set storage unit 332 stores a word set of attribute values of the character string type attribute created by the attribute value word extraction unit 331.

  The character string type attribute similarity calculation unit 333 includes two character string type attributes constituting different data tables based on the word set of attribute values of the character string type attributes stored in the attribute value word set storage unit 332. The similarity between them is calculated. Specifically, the character string type attribute similarity calculation unit 333 calculates the two characters based on the number of words that match between the word sets of the attribute values of the two character string type attributes that form different data tables. Calculate the similarity between column type attributes. Note that the character string type attribute similarity calculation unit 333 calculates the similarity for all combinations of two character string type attributes constituting different data tables as described above. The similarity calculated by the character string type attribute similarity calculating unit 333 as described above is stored in the character string type attribute similarity storing unit 26.

  Here, a plurality of different data tables stored in the data table storage unit 22 shown in FIG. 2 will be described with reference to FIGS.

  In the present embodiment, for example, it is assumed that analysis of business errors (related data) in a plurality of banks is supported. Here, in the plurality of data tables stored in the data table storage unit 22, for example, data (information) relating to mistakes (such as wrong commissions, wrong account number designations, etc.) that occurred in daily operations of banks A to C is accumulated. The data table for each bank (that is, the data table of banks A to C).

  FIG. 5 shows an example of the data structure of the bank A data table among the plurality of data tables stored in the data table storage unit 22.

  As shown in FIG. 5, the bank A data table 221 has attribute names (attribute names) of “version”, “occurrence date”, “discovery date”, “occurrence cause / occurrence person”, “phenomenon / occurrence person”. ”,“ Occurrence work ”,“ occurrence position ”,“ loss amount (yen) ”,“ occurrence store number ”, and“ discovered store number ”. In the following description, for example, an attribute whose attribute name is “version” is simply referred to as a “version” attribute. The other attributes are also expressed in the same manner.

  Here, among the plurality of attributes constituting the bank A data table 221 shown in FIG. 5, for example, the “loss amount (yen)” attribute is “94500”, “300000”, “1500000”, “0”, and the like. Has an attribute value that contains a numeric value. Therefore, the “loss amount (yen)” attribute is a numerical attribute. The “version” attribute, the “occurrence date” attribute, the “discovery date” attribute, the “loss amount (yen)” attribute, the “occurrence store number” attribute, The “discovered store number” attribute is a numeric attribute.

  Further, among the plurality of attributes constituting the bank A data table 221 shown in FIG. 5, for example, the “occurrence cause / occurrence” attribute includes “insufficient experience”, “insufficient instruction / education”, “accident by third party” ”And attribute values including character strings such as“ customer request error / entry error ”. Therefore, the “occurrence cause / occurrence” attribute is a character string type attribute. Of the plurality of attributes constituting the bank A data table 221, the “occurrence / occurrence” attribute, the “phenomenon / occurrence” attribute, the “occurrence work” attribute, and the “occurrence position” attribute are character string types. Attribute.

  FIG. 6 shows an example of the data structure of the B bank data table among the plurality of data tables stored in the data table storage unit 22.

  As shown in FIG. 6, the data table 222 of the bank B includes an “occurrence date” attribute, a “version” attribute, a “found date” attribute, an “occurrence cause” attribute, an “overview” attribute, and a “job title / occurrence” attribute. , “Position / verifier” attribute, “direct loss (thousand yen)” attribute, “indirect loss (thousand yen)” attribute, “business” attribute, and “occurring store” attribute. Since the bank B data table 222 is defined differently from the bank A data table 221 described above, the attribute name and the attribute value notation are compared with the attributes constituting the bank A data table 221. Is different.

  Here, among the plurality of attributes constituting the bank B data table 222 shown in FIG. 6, for example, the “indirect loss amount (thousand yen)” attribute is “0”, “400”, “0”, and “0”. Attribute values including numerical values such as Therefore, the “indirect loss amount (thousand yen)” is a numerical attribute. The “occurrence date” attribute, the “version” attribute, the “identification date” attribute, the “direct loss (1,000 yen)” attribute, the “indirect loss ( The “1,000 yen)” attribute and the “occurring store” attribute are numeric attributes.

  Further, among the plurality of attributes constituting the bank B data table 222 shown in FIG. 6, for example, the “occurrence cause” attributes are “knowledge / experience / learning shortage”, “accident by customer”, “complex work contents”. And an attribute value including a character string such as “careless mistake”. For this reason, the “occurrence cause” attribute is a character string type attribute. Of the plurality of attributes constituting the bank B data table 222, the "occurrence cause" attribute, the "summary" attribute, the "position / issuer" attribute, the "position / verifier" attribute, and the "business" attribute are characters. It is a column type attribute.

  FIG. 7 shows an example of the data structure of the C bank data table among the plurality of data tables stored in the data table storage unit 22.

  As shown in FIG. 7, the data table 223 of bank C includes a “found date” attribute, “occurrence date” attribute, “version” attribute, “occurrence store number” attribute, “occurrence cause / occurrence” attribute, “phenomenon” ”Attribute,“ occurrence work ”attribute,“ occurrence position ”attribute,“ loss (thousand yen) ”attribute, and“ risk evaluation ”attribute. Since the bank C data table 223 is defined differently from the bank A data table 221 and the bank B data table 222 described above, the bank A data table 221 and the bank B data table 222 are configured. The attribute name and attribute value are different from each other.

  Here, among the plurality of attributes constituting the bank C data table 223 shown in FIG. 7, for example, the “risk evaluation” attribute includes numerical values such as “1”, “0”, “3”, and “0”. Has an attribute value. For this reason, the “risk evaluation” attribute is a numerical attribute. Of the plurality of attributes constituting the bank C data table 223, the “found date” attribute, the “occurrence date” attribute, the “version” attribute, the “occurrence store number” attribute, and the “loss (thousand yen)” attribute And the “risk assessment” attribute is a numeric attribute.

  Among the plurality of attributes constituting the bank C data table 223 shown in FIG. 7, for example, the “generator position” attribute includes “general employee”, “senior employee”, “general employee”, “part”, and the like. Has an attribute value that contains a string. For this reason, the “generation position” attribute is a character string type attribute. Of the plurality of attributes constituting the bank C data table 223, the “cause / occurrence” attribute, the “phenomenon” attribute, the “occurrence work” attribute, and the “occurrence position” attribute are character string type attributes. .

  Next, the processing procedure of the data analysis support apparatus 30 according to the present embodiment will be described with reference to the flowchart of FIG.

  First, the attribute type classification unit 31 acquires a plurality of data tables stored in the data table storage unit 22 (step S1).

  Next, the attribute type classification unit 31 classifies each attribute constituting the data table into a numeric type attribute or a character string type attribute for each acquired data table (step S2). Further, the attribute type classification unit 31 further classifies each of the attributes classified as numerical type attributes into the numerical type attribute small classification. This numerical attribute small classification includes, for example, an integer type, a floating point type, and a date type.

  The attribute type classification unit 31 refers to the attribute value of the attribute that constitutes the acquired data table and executes the attribute classification process. In addition, when the information (type information) specifying each type of the attribute constituting the data table acquired by the attribute type classification unit 31 is held in the data table, the classification process is performed with reference to the information. May be executed.

  The attribute type classification unit 31 stores the classification results in the numerical type attribute storage unit 23 and the character string type attribute storage unit 24 for each data table (step S3).

  Here, it is assumed that the plurality of data tables stored in the data table storage unit 22 are the above-described data tables 221 to 223 of the A to C banks.

  In this case, the numeric attribute storage unit 23 stores numeric attributes for each of the data tables 221 to 223 of the A to C banks. Specifically, the numeric type attribute storage unit 23 includes a numeric type attribute (a numeric type attribute constituting the bank A data table 221) among a plurality of attributes constituting the bank A data table 221; A numerical type attribute (a numerical type attribute constituting the bank B data table 222) among a plurality of attributes constituting the data table 222 and a numeric type attribute (C for a plurality of attributes constituting the data table 223 for the C bank) (Numeric type attributes constituting the bank data table 223) are stored.

  In the numeric attribute storage unit 23, the numeric attributes constituting the bank A data table 221 are further classified into an integer type, a floating point type, and a date type as described above. The same applies to the numeric type attributes constituting the bank B data table 222 and the numeric type attributes constituting the bank C data table 223.

  The numerical type attributes of the bank data tables 221 to 223 stored in the numerical type attribute storage unit 23 include the attribute values of the numerical type attributes in the data tables 221 to 223 of the bank A to C and The attribute name of the numeric attribute is included.

  On the other hand, the character string type attribute storage unit 24 stores a character string type attribute for each of the data tables 221 to 223 of the A to C banks. Specifically, the character string type attribute storage unit 24 includes a character string type attribute (character string type attribute constituting the bank A data table 221) among a plurality of attributes constituting the bank A data table 221; Among the plurality of attributes constituting the bank B data table 222, the character string type attribute (character string type attribute constituting the bank B data table 222) and the plurality of attributes constituting the bank C data table 223 A character string type attribute (character string type attribute constituting the data table 223 of the C bank) is stored.

  Further, the character string type attributes of the data tables 221 to 223 of the A to C banks stored in the character string type attribute storage unit 24 have the character string type attributes in the data tables 221 to 223 of the A to C banks. The attribute value and the attribute name of the character string type attribute are included.

  Next, the similarity calculation process is executed with reference to the numerical attribute storage unit 23 and the character string attribute storage unit 24 (step S4). Although details will be described later, the similarity calculation processing includes numerical value attribute similarity calculation processing executed by the numerical attribute processing unit 32 and character string attribute processing executed by the character string attribute processing unit 33. Similarity calculation processing is included.

  In the numerical value attribute similarity calculation process, based on attribute values (numerical values included in) two numerical type attributes that constitute different data tables (each) stored in the numerical type attribute storage unit 23, The similarity between the two numeric type attributes is calculated by the numeric type attribute processing unit 32. In the numerical attribute similarity calculation process, the similarity between two numerical attributes having the same numerical attribute small classification (that is, classification destination) is calculated. In the numerical value attribute similarity calculation process, two numerical type attributes that are different numerical tables stored in the numerical type attribute storage unit 23 and that have the same numerical type attribute sub-classification are included. Similarities are calculated for all combinations.

  Further, in the similarity calculation processing of the character string type attribute, based on the attribute values (character strings included in) the two character string type attributes constituting the different data tables stored in the character string type attribute storage unit 24. Thus, the similarity between the two character string type attributes is calculated by the character string type attribute processing unit 33. In this character string type attribute similarity calculation process, the similarity is calculated for all combinations of two character string type attributes constituting different data tables stored in the character string type attribute storage unit 24.

  When the processing of step S4 is executed, the numerical attribute similarity storage unit 25 stores the similarity calculated by the numerical attribute processor 32 (similarity for each combination of two numerical attributes constituting different data tables). The similarity list including the degree of similarity is stored, and the character string type attribute similarity storage unit 26 stores the similarity calculated by the character string type attribute processing unit 33 (two character string type attributes constituting different data tables). (Similarity for each combination) is stored (step S5). Details of the similarity list stored in the numerical attribute similarity storage unit 25 and the character string attribute similarity storage unit 25 will be described later.

  Here, the threshold value input unit 34 inputs a threshold value specified by the user, for example (step S6). Here, the threshold value has been described as being specified by the user. However, the threshold value may be set in advance in the data analysis support device 30, for example, or dynamically determined according to the contents of the data table or the like. It does not matter.

  Next, the similarity attribute candidate extraction unit 35 includes a similarity list stored in the numerical attribute similarity storage unit 25, a similarity list stored in the character string attribute similarity storage unit 26, and a threshold input unit 34. A process for extracting similar attribute candidates (hereinafter referred to as a similar attribute candidate extraction process) is executed based on the threshold value input by (step S7). The similar attribute candidate extracted by the similar attribute candidate extraction unit 35 in the similar attribute candidate extraction process includes a combination of two numerical type attributes (numerical type attribute pair) and different data tables constituting the different data tables described above. A combination of two character string type attributes (character string type attribute pair) is included. The details of the similar attribute candidate extraction process will be described later.

  When the process of step S7 is executed, the similar attribute candidates extracted by the similar attribute candidate extraction unit 35 are stored in the similar attribute candidate storage unit 27 (step S8). Note that the similar attribute candidates stored in the similar attribute candidate storage unit 27 are attribute candidates that represent the same content among a plurality of different data tables stored in the data table storage unit 22. It can be used when comparing data tables.

  Next, with reference to the flowchart of FIG. 9, the processing procedure of the similarity calculation process of the numerical attribute included in the above-described similarity calculation process (the process of step S4 shown in FIG. 8) will be described. The numerical value attribute similarity calculation processing is executed by the numerical value attribute processing unit 32.

  First, the numerical type attribute processing unit 32 executes the following steps S11 and S12 for each of the numerical type attributes stored in the numerical type attribute storage unit 23. Here, the numerical type attribute that is the target of this processing is referred to as a target numerical type attribute.

  The attribute value numerical value range specifying unit 321 included in the numerical value type attribute processing unit 32 specifies a numerical value range of attribute values of the target numerical value attribute stored in the numerical value attribute storage unit 23 (step S11). In this case, the attribute value numerical value range specifying unit 321 specifies the maximum value and the minimum value of the attribute values (the numerical values included in) the target numerical value attribute stored in the numerical value attribute storage unit 23, and from the maximum value The range up to the minimum value is the numeric range of the attribute value of the target numeric type attribute.

  The attribute value numerical range specifying unit 321 stores the numerical value range of the attribute value of the specified target numerical value attribute in the attribute value numerical range storage unit 322 (step S12).

  Here, it is determined whether or not the processes of steps S11 and S12 described above have been executed for all the numeric type attributes stored in the numeric type attribute storage unit 23 (step S13).

  If it is determined that processing has not been executed for all the numeric type attributes stored in the numeric type attribute storage unit 23 (NO in step S13), the process returns to step S11 described above and is repeated. In this case, the process is executed with the numeric type attribute for which the processes of steps S11 and S12 have not been executed as the target numeric type attribute. As described above, the processing of steps S11 and S12 is executed for all the numeric type attributes stored in the numeric type attribute storage unit 23, whereby the attribute value of the numeric type attribute specified for each numeric type attribute is included. Are stored in the attribute value numerical value range storage unit 322. Hereinafter, the numerical value range of the attribute value included in the numerical value attribute stored in the attribute value numerical value range storage unit 322 is simply referred to as the numerical value range of the numerical value attribute.

  On the other hand, when it is determined that the processing has been executed for all the numeric type attributes stored in the numeric type attribute storage unit 23 (YES in step S13), the numeric type attribute similarity calculation unit 323 is The following steps S14 and S15 are executed for each of the combinations of two numeric type attributes (numeric type attribute pairs) constituting different data tables stored in. Here, the numerical attribute pair that is the target of this processing is referred to as a target numerical attribute pair. One numerical value attribute included in the target numerical value attribute pair is referred to as a first numerical value attribute, and the other numerical value attribute is referred to as a second numerical value attribute. It is assumed that the first numerical value attribute and the second numerical value attribute have the same numerical attribute small classification (that is, classification destination).

  First, the numerical value attribute similarity calculation unit 323 acquires the numerical value ranges of the first and second numerical value attributes included in the target numerical value attribute pair from the attribute value numerical value storage unit 322.

  Next, the numerical value attribute similarity calculation unit 323 determines the similarity (target numerical value type) between the first and second numerical value attributes based on the acquired numerical value ranges of the first and second numerical value attributes. The similarity of the attribute pair is calculated (step S14).

  Here, it is assumed that the numerical value range of the first numerical value attribute acquired by the numerical value attribute similarity calculation unit 323 is wider than the numerical value range of the second numerical value attribute, or the numerical value ranges of both are equal. . In this case, the numerical value attribute similarity calculation unit 323 determines the ratio (ratio) of the overlapping range of the numerical ranges of the first and second numerical attribute to the numerical range of the first numerical attribute. Calculated as the similarity between two numeric attributes.

  On the other hand, when the numerical value range of the first numerical value attribute is narrower than the numerical value range of the second numerical value attribute, the numerical value attribute similarity calculation unit 323 performs the first and second values for the numerical value range of the second numerical value attribute. The ratio (ratio) of the overlapping range of the numerical ranges of the numerical type attributes is calculated as the similarity between the first and second numerical type attributes.

  Here, with reference to FIG. 10 and FIG. 11, the similarity between the first and second numerical type attributes calculated by the numerical type attribute similarity calculating unit 323 will be specifically described.

  First, in FIG. 10, the first numeric type attribute is the “loss amount (yen)” attribute that constitutes the bank A data table 221, and the second numeric type attribute is the bank B data table 222. It is assumed that the attribute is “direct loss (1,000 yen)”. Further, the numerical range of the first numerical type attribute (that is, the “loss amount (yen)” attribute) is 0 to 1500000, and the second numerical type attribute (that is, the “direct loss amount (thousand yen)” attribute). The numerical range of is assumed to be 0 to 1000000. In the first and second numerical type attributes, the unit is different between “yen” and “thousand yen”. For example, when the numerical range is specified as described above, the unit is the same. It shall be corrected.

  Here, the numerical range of the first numerical attribute is wider than the numerical range of the second numerical attribute. In this case, the similarity between the first and second numeric type attributes overlaps the numeric range of the first and second numeric type attributes with respect to the numeric range of the first numeric type attribute (here, 0 to 1500000). The ratio of the range (here, 0 to 1000000), that is, 1000000 / 1500,000≈0.667 is calculated.

  On the other hand, in FIG. 11, the first numeric type attribute is the “loss amount (yen)” attribute constituting the bank A data table 221, and the second numeric type attribute constitutes the bank B data table 222. It is assumed that the attribute is “occurring store”. The numerical range of the first numerical type attribute (that is, the “loss amount (yen)” attribute) is 0 to 1500000, and the numerical range of the second numerical type attribute (that is, the “occurring store” attribute) is 1. ˜145.

  Here, the numerical range of the first numerical attribute is wider than the numerical range of the second numerical attribute. In this case, the similarity between the first and second numeric type attributes overlaps the numeric range of the first and second numeric type attributes with respect to the numeric range of the first numeric type attribute (here, 0 to 1500000). The ratio of the range (here, 1 to 145), that is, 144 / 1500000≈0.000 is calculated.

  Returning to FIG. 9 again, the numerical attribute similarity calculation unit 323 stores the calculated similarity between the first and second numerical attributes in the similarity list (step S15). In this case, the degree of similarity between the first and second numeric type attributes is the combination of the data table constituted by the first numeric type attribute and the data table constituted by the second numeric type attribute. It is stored in the prepared similarity degree list of the numerical type attribute small classification of the first and second numerical type attributes. Specifically, the similarity between the integer type first numeric type attribute constituting the bank A data table 221 and the integer type second numeric type attribute constituting the bank B data table 222 is determined by the bank. The data is stored in an integer similarity list prepared for the combination of the A data table 221 and the bank B data table 222.

  Note that the similarity list is prepared for each combination of two different data tables constituting each of two numerical type attributes (that is, the first and second numerical type attributes) for which the similarity is calculated. . A similarity list prepared for each combination of two different data tables is further prepared for each numerical subclass.

  For example, when the number of data tables stored in the data table storage unit 22 is n and the number of numerical type attribute subclasses into which the numerical type attribute is classified is m, n * (n−1) *. A maturity number list of m / 2 is prepared. Specifically, for example, in the case where three data tables 221 to 223 are stored in the data table storage unit 22 and there are three numerical type attribute minor classifications (for example, an integer type, a floating point type, and a date type). Nine similarity degree lists are prepared.

  Next, it is determined whether or not the processes of steps S14 and S15 described above have been executed for all combinations of two numeric type attributes constituting different data tables (that is, all numeric type attribute pairs) (step S16). ).

  When it is determined that the processing has not been executed for all the numeric attribute pairs (NO in step SS16), the processing returns to the above-described step S14 and is repeated. In this case, the process is executed with the numeric attribute pair that has not been subjected to the processes of steps S14 and S15 as the target numeric attribute pair.

  On the other hand, when it is determined that the processing has been executed for all the numerical attribute pairs (YES in step S16), the numerical attribute similarity calculation processing ends.

  As described above, when the numerical attribute similarity calculation process is executed, all prepared similarity lists (numerical attribute similarity lists) are created. The similarity list created in the numerical attribute similarity calculation process is stored in the numerical attribute similarity storage unit 25 as described above.

  Note that the method of calculating the degree of similarity between two numeric attributes constituting different data tables is not limited to the above. For example, a basic statistic such as an average or a variance is compared, or a graph is created to calculate the similarity. Other methods such as making the overlap between graphs similar may be used.

  Here, with reference to FIG. 12, the similarity list created in the numerical value attribute similarity calculation processing will be specifically described. FIG. 12 shows an example of an integer type similarity list prepared for the combination of the bank A data table 221 and the bank B data table 222.

  As shown in FIG. 12, in the similarity list 251, each of the numeric type attribute that constitutes the bank A data table 221 and the numeric type attribute that constitutes the bank B data table 222 is shown. The numerical type attributes constituting the bank A data table 221 include a “version” attribute, a “loss amount (yen)” attribute, an “occurrence store number” attribute, and a “discovered store number” attribute. The “version” attribute, the “loss amount (yen)” attribute, the “occurrence store number” attribute, and the “discovered store number” attribute are integer type numeric attributes. In addition, the numerical type attributes constituting the bank B data table 222 include a “version” attribute, a “direct loss (thousand yen)” attribute, an “indirect loss (thousand yen)” attribute, and an “occurring store” attribute. included. Similarly, the “version” attribute, the “direct loss (thousand yen)” attribute, the “indirect loss (thousand yen)” attribute, and the “occurring store” attribute are integer type numeric attributes.

  In the example illustrated in FIG. 12, the similarity list 251 stores 0.805 in association with, for example, the “version” attribute and the “version” attribute. According to this, the similarity between the “version” attribute and the “version” attribute constituting the different data tables (here, the bank A data table 221 and the bank B data table 222) is 0.805. It is shown.

  In the similarity list 251 shown in FIG. 12, similarities between two numerical type attributes other than the “version” attribute and the “version” attribute are similarly stored. In other words, the similarity list 251 stores similarities for all combinations of the integer numeric attributes constituting the bank A data table 221 and the integer numeric attributes constituting the bank B data table 222. Has been.

  Here, the integer type similarity list prepared for the combination of the bank A data table 221 and the bank B data table 222 has been described, but the same applies to other similarity lists. Detailed description is omitted.

  Next, with reference to the flowchart of FIG. 13, a processing procedure of the similarity calculation process of the character string type attribute included in the above-described similarity calculation process (the process of step S4 shown in FIG. 8) will be described. Note that the character string type attribute similarity calculation processing is executed by the character string type attribute processing unit 33.

  First, the character string type attribute processing unit 33 performs the following steps S21 and S22 for each of the character string type attributes stored in the character string type attribute storage unit 24. Here, the character string type attribute to be processed is referred to as a target character string type attribute.

  The attribute value word extraction unit 331 included in the character string type attribute processing unit 33 performs morphological analysis on the character string included in the attribute value of the target character string type attribute stored in the character string type attribute storage unit 24. As a result, the attribute value word extraction unit 331 extracts words constituting the character string included in the attribute value of the target character string type attribute, and a word set including the extracted words (hereinafter, the target character string type attribute). Are created (step S21).

  Here, with reference to FIG. 14, the word set of the object character string type | mold attribute produced by the attribute value word extraction part 331 is demonstrated concretely. Here, it is assumed that the target character string type attribute is an “occurrence cause / occurrence person” attribute constituting the data table 221 of the A bank shown in FIG.

  First, a set of all attribute values of the target character string type attribute (hereinafter referred to as an attribute value set of the target character string type attribute) in the data table configured by the target character string type attribute is acquired. In the example shown in FIG. 14, the attribute value set of the target string type attribute includes attribute values “insufficient experience”, “insufficient instruction / education”, “accident by third party”, “insufficient instruction / education”, “customer "Incorrect request / entry error", "insufficient experience", "insufficient instruction / education", "insufficient instruction / education", and "accidents by third parties".

  Next, by collecting the same attribute values among the attribute values included in the attribute value set of the target string type attribute into one (that is, excluding duplicates), the unique attribute value set of the target string type attribute Is created (step S31). In the example shown in FIG. 14, the unique attribute value set of the target string type attribute includes attribute values “insufficient experience”, “insufficient instruction / education”, “accident by third party”, and “customer request error / entry "Error" is included. The attribute value set of the target string type attribute described above contains multiple attribute values “insufficient experience”, “insufficient teaching / education”, and “accidents by third parties”. The column type attributes are aggregated into one in a unique attribute value set.

  Next, the character string is divided into words by performing morphological analysis on the attribute value (character string included) included in the unique attribute value set of the target character string type attribute. Here, the morpheme analysis process is a process of dividing a character string into words and assigning parts of speech to the divided words. Specifically, the attribute value “Customer's request mistake / entry error” is converted into “O (prefix) / customer (noun) / like (suffix) / no (particle) / request (noun) by morphological analysis processing. / Miss (noun) /. (Symbol) / entry (noun) / error (noun) ”. A word whose part of speech is a noun is extracted from such a morphological analysis processing result, and a word set including the word is created (step S32). In the example illustrated in FIG. 14, the word set created by the attribute value word extraction unit 331 includes the words “experience”, “insufficiency”, “guidance”, “education”, “insufficiency”, “third party”. , “Accident”, “customer”, “request”, “miss”, “entry” and “error” are included.

  In the process of step S32, in addition to the word whose part of speech is a noun, for example, a word whose part of speech is a verb, an unknown word, and the like may be extracted together. An unknown word is a word which is not registered in the dictionary for morphological analysis, for example. In general, there is a high possibility that proper nouns or technical terms become unknown words.

  Next, by collecting the same words among the words included in the created word set into one (that is, excluding duplication), a unique word set (attribute value word set) of the target string type attribute is obtained. It is created (step S33). In the example shown in FIG. 14, the unique word set of the target character string type attribute includes the words “experience”, “insufficiency”, “guidance”, “education”, “third party”, “accident”, “customer”. , “Request”, “miss”, “entry”, and “error”. Since the word set created in step S32 described above includes a plurality of words “insufficiency”, these words are integrated into one in the unique word set of the target character string type attribute.

  Returning to FIG. 13 again, the attribute value word extracting unit 331 uses the attribute value word set storage unit 332 as a target character string type attribute word set (unique word set of the target character string type attribute) created as described above. (Step S22).

  Here, it is determined whether or not the processing of steps S21 and S22 described above has been executed for all the character string type attributes stored in the character string type attribute storage unit 24 (step S23).

  When it is determined that processing has not been performed for all the character string type attributes stored in the character string type attribute storage unit 24 (NO in step S23), the process returns to the above step S21 and is repeated. In this case, the processing is executed with the character string type attribute for which the processing of steps S21 and S22 has not been executed as the target character string type attribute. As described above, by executing the processing of steps S21 and S22 for all the character string type attributes stored in the character string type attribute storage unit 24, the word set for each character string type attribute is stored in the attribute value word set. Stored in the unit 332.

  On the other hand, if it is determined that the processing has been executed for all the character string type attributes stored in the character string type attribute storage unit 24 (YES in step S23), the character string type attribute similarity calculation unit 333 displays the character string type attribute value. The following steps S24 to S26 are executed for each combination of two character string type attributes (character string type attribute pair) constituting different data tables stored in the attribute storage unit 24. Here, the character string type attribute pair that is the target of this processing is referred to as a target character string type attribute pair. One character string type attribute included in the target character string type attribute pair is referred to as a first character string type attribute, and the other character string type attribute is referred to as a second character string type attribute.

  First, the character string type attribute similarity calculation unit 333 acquires the word set of the first and second character string type attributes included in the target character string type attribute pair from the attribute value word set storage unit 332.

  Next, the character string type attribute similarity calculation unit 333 refers to the acquired word sets of the first and second character string type attributes, and determines between the word sets of the first and second character string type attributes. The number of matching words is specified (step S24). In this case, the character string type attribute similarity calculation unit 333 selects a word that matches a word included in the word set of the second character string type attribute among words included in the word set of the first character string type attribute. The number (hereinafter referred to as the number of matches of the first character string type attribute) is specified. Further, the character string type attribute similarity calculating unit 333 counts the number of words that match the word included in the word set of the first character string type attribute among the words included in the word set of the second character string type attribute. (Hereinafter referred to as the number of matches of the second character string type attribute).

  The character string type attribute similarity calculation unit 333 calculates the degree of similarity between the first and second character string type attributes based on the number of matches between the identified first and second character string type attributes ( Step S25). In this case, the character string type attribute similarity calculation unit 333 performs matching of words included in the word set of the first character string type attribute (hereinafter referred to as word matching rate of the first character string type attribute) and the first The similarity is calculated using the matching rate of words included in the word set of the second character string type attribute (hereinafter referred to as the word matching rate of the second character string type attribute).

  Here, the word match rate of the first character string type attribute is the first character string type attribute specified in step S24 described above with respect to the number of words included in the word set of the first character string type attribute. Refers to the percentage of matches. The word match rate of the second character string type attribute is the match of the second character string type attribute specified in step S24 described above with respect to the number of words included in the word set of the second character string type attribute. The ratio of numbers.

  In this case, the character string type attribute similarity calculating unit 333 calculates the average value of the word match rates of the first and second character string type attributes as the similarity between the first and second character string type attributes. To do.

  Here, with reference to FIG. 15, the similarity between the 1st and 2nd character string type | mold attributes calculated by the character string type | mold attribute similarity calculation part 333 is demonstrated concretely.

  Here, the first character string type attribute is the “cause / occurrence” attribute that constitutes the bank A data table 221, and the second character string type attribute is the “occurrence” that constitutes the bank B data table 222. Attribute.

  As shown in FIG. 15, the word set of the first string type attribute (that is, the “cause / occurrence” attribute) has 11 words, specifically, the words “experience” and “insufficient”. , “Guidance”, “education”, “third party”, “accident”, “customer”, “request”, “miss”, “entry” and “error”. The word set of the second character string type attribute (that is, the “occurrence cause” attribute) has 10 words, specifically, the words “knowledge”, “experience”, “education”, “insufficient”. , “Customer”, “accident”, “complexity”, “work”, “content” and “careless mistake”.

  Here, among words included in the word set of the first character string type attribute, words that match the words included in the word set of the second character string type attribute are words “experience”, “insufficient”, “ “Education” and “Accident”. In this case, the number of matches of the first character string type attribute is 4. On the other hand, of the words included in the word set of the second string type attribute, the words that match the words included in the word set of the first string type attribute are the words “experience”, “education”, “insufficient” "And" accidents ". In this case, the number of matches of the second character string type attribute is 4.

  Moreover, since the number of words included in the word set of the first character string type attribute is 11 as described above, the word matching rate of the first character string type attribute is 4/11. Further, since the number of words included in the word set of the second character string type attribute is 10, the word matching rate of the second character string type attribute is 4/10.

  Accordingly, the similarity between the first and second character string type attributes is calculated as an average value of 4/11 and 4/10, that is, (4/11 + 4/10) /2≈0.382. .

  Returning to FIG. 13 again, the character string type attribute similarity calculating unit 333 stores the calculated similarity between the first and second character string type attributes in the similarity list (step S26). In this case, the similarity between the first and second character string type attributes is a combination of a data table constituted by the first character string type attribute and a data table constituted by the second character string type attribute. Is stored in the similarity list prepared for. Specifically, the similarity between the first character string type attribute constituting the bank A data table 221 and the second character string type attribute constituting the bank B data table 222 is determined by the bank A data table. 221 and the data table 222 of the bank B are stored in the similarity list prepared.

  Note that the similarity list is prepared for each combination of two different data tables constituting each of the two character string type attributes (that is, the first and second character string type attributes) whose similarity is calculated. ing.

  For example, when the number of data tables stored in the data table storage unit 22 is n, n * (n−1) / 2 number of similarity lists are prepared. Specifically, for example, when three data tables 221 to 223 are stored in the data table storage unit 22, three similarity list tables are prepared.

  Next, it is determined whether or not the processing in steps S24 to S26 described above has been executed for all combinations of two character string type attributes constituting different data tables (that is, all character string type attribute pairs) ( Step S27).

  When it is determined that the processing has not been executed for all the character string type attribute pairs (NO in step S27), the processing returns to the above-described step S24 and is repeated. In this case, the process is executed with the character string type attribute pair that has not been subjected to the processes of steps S24 to S26 as the target character string type attribute pair.

  On the other hand, when it is determined that the process has been executed for all the character string type attribute pairs (YES in step S27), the character string type attribute similarity calculation process ends.

  As described above, when the character string type attribute similarity calculation process is executed, all prepared similarity degree lists (similarity list of character string type attributes) are created. The similarity list created in the character string attribute similarity calculation processing is stored in the character string attribute similarity storage unit 26 as described above.

  Here, the similarity list created in the similarity calculation process of the character string type attribute will be specifically described with reference to FIG. FIG. 16 shows an example of a similarity list prepared for a combination of the bank A data table 221 and the bank B data table 222.

  As shown in FIG. 16, the similarity list 261 shows character string type attributes constituting the bank A data table 221 and string type attributes constituting the bank B data table 222. . The character string type attributes constituting the bank A data table 221 include a “cause / occurrence” attribute, a “phenomenon / occurrence” attribute, an “occurrence work” attribute, and an “occurrence position” attribute. Further, the character string type attributes constituting the bank B data table 222 include an “occurrence cause” attribute, an “overview” attribute, a “position / occurrence” attribute, and a “position / verifier” attribute.

  In the example shown in FIG. 16, the similarity list 261 stores 0.382 in association with, for example, the “occurrence / occurrence” attribute and the “occurrence cause” attribute. According to this, the similarity between the “occurrence cause / occurrence” attribute and the “occurrence cause” attribute constituting different data tables (here, the data table 221 of the A bank and the data table 222 of the B bank) is 0. 382.

  The similarity list 261 shown in FIG. 16 similarly stores similarities between two other character string type attributes other than the “cause / occurrence” attribute and the “cause” attribute. . That is, the similarity list 261 stores the degrees of similarity for all combinations of the character string type attributes constituting the bank A data table 221 and the character string type attributes constituting the bank B data table 222.

  Here, the similarity list prepared for the combination of the bank A data table 221 and the bank B data table 222 has been described, but the same applies to the other similarity lists, so a detailed description thereof will be given. Omitted.

  Next, the procedure of the similar attribute candidate extraction process (the process of step S7 shown in FIG. 8) described above will be described with reference to the flowchart of FIG. This similar attribute candidate extraction process is a process executed in common for the numeric type attribute and the character string type attribute. This similar attribute candidate extraction process is executed by the similar attribute candidate extraction unit 35.

  The similar attribute candidate extraction unit 35 executes the following steps S41 to S49 for each of the similarity list stored in the numerical attribute similarity storage unit 25 and the character string attribute similarity storage unit 26. Here, the similarity list targeted for this processing is referred to as a target similarity list.

  First, the similar attribute candidate extraction unit 35 extracts the target similarity list from the numerical attribute similarity storage unit 25 or the character string attribute similarity storage unit 26 (step S41).

  Next, the similar attribute candidate extraction unit 35 performs the following steps S42 to S48 for each of the attributes in the acquired target similarity list. Here, the attribute to be processed is referred to as a target attribute.

  The similar attribute candidate extraction unit 35 extracts a target attribute from the target similarity list (step S42).

  In the target similarity list, the similar attribute candidate extraction unit 35 has a similarity with the extracted target attribute equal to or greater than the threshold input by the threshold input unit 34 (threshold input in step S6 shown in FIG. 8). Search for attributes (attributes constituting a data table different from the target attribute). This threshold is a value indicating a degree of similarity sufficient for extracting similar attribute candidates.

  In the following description, an attribute whose similarity to the target attribute is equal to or greater than a threshold value is referred to as a corresponding attribute.

  The similar attribute candidate extraction unit 35 determines whether or not the number of searched corresponding attributes is two or more (step S43).

  When it is determined that the number of corresponding attributes is not two or more (NO in step S43), the similar attribute candidate extraction unit 35 determines whether or not the number of searched corresponding attributes is one (step S44). ).

  When it is determined that the number of corresponding attributes is not one (that is, the corresponding attribute does not exist) (NO in step S44), the similar attribute candidate extracting unit 35 is different from the target attribute in the target similarity list. It is determined whether or not there is an attribute satisfying a predetermined condition among the attributes constituting the table (step S45). Here, the predetermined condition includes the presence of an attribute having a maximum similarity to the target attribute as compared to other attributes.

  Here, the process for determining whether or not there is an attribute having the maximum similarity to the symmetry attribute (that is, the process of step S45) will be specifically described. Here, it is assumed that the target similarity list is the above-described similarity list 261 shown in FIG. The threshold value input by the threshold value input unit 34 is assumed to be 0.8.

  First, it is assumed that the target attribute is a “phenomenon / occurrence” attribute that configures the data table 221 of the bank A in the similarity list 261. In this case, in the similarity list 261, there is no attribute (corresponding attribute) whose similarity with the “phenomenon / occurrence” attribute that is the target attribute is greater than or equal to the threshold (0.8), and therefore the process of step S45 is Executed.

  Referring to the similarity list 261 shown in FIG. 16, in the similarity list 261, the attribute having the maximum similarity with the “phenomenon / occurrence” attribute that is the target attribute is the “summary” attribute.

  On the other hand, in the similarity list 261, the attribute having the maximum similarity with the “summary” attribute is the “phenomenon / occurrence” attribute that is the target attribute.

  In this way, the attribute having the maximum similarity with the “phenomenon / occurrence” attribute as the target attribute is the “summary” attribute, and the attribute having the maximum similarity with the “summary” attribute is “phenomenon / occurrence”. In the case of an attribute (that is, a target attribute), it is determined that there is an attribute (here, “summary” attribute) having a maximum similarity to the target attribute.

  Next, it is assumed that the target attribute is an “occurring business” attribute that constitutes the data table 221 of the bank A in the similarity list 261. In this case, in the similarity list 261, since there is no attribute (corresponding attribute) whose similarity with the “occurring work” attribute that is the target attribute is equal to or greater than the threshold (0.8), the process of step S45 is executed. The

  Referring to the similarity list 261 shown in FIG. 16, in the similarity list 261, the attribute having the highest similarity with the “occurring business” attribute that is the target attribute is the “cause” attribute.

  On the other hand, in the similarity list 261, the attribute having the highest similarity to the “occurrence cause” attribute is the “occurrence cause / occurrence” attribute, not the “occurrence work” attribute that is the target attribute.

  In this way, the attribute that has the maximum similarity to the “occurrence work” attribute that is the target attribute is the “occurrence cause” attribute, but the attribute that has the maximum similarity to the “occurrence cause” attribute is “occurrence work”. If it is not an attribute (that is, a target attribute), it is determined that there is no attribute having the maximum similarity with the target attribute.

  As described above, in step S45, when the attribute having the maximum similarity with the target attribute and the attribute having the maximum similarity are the target attributes, there is an attribute having the maximum similarity with the target attribute. If the attribute having the maximum similarity with the target attribute and the attribute having the maximum similarity are not the target attribute, there is no attribute having the maximum similarity with the target attribute (that is, the similar attribute No candidate).

  If it is determined that there is an attribute having the maximum degree of similarity with the target attribute, that is, it is determined that there is an attribute that satisfies a predetermined condition (YES in step S45), the attribute and the target attribute (combination) are similar Extracted as attribute candidates (step S46).

  On the other hand, when it is determined that there is no attribute having the maximum similarity with the target attribute, that is, there is no attribute satisfying a predetermined condition (NO in step S45), the process in step S46 is not executed.

  If it is determined in step S43 that the number of corresponding attributes is two or more, the similar attribute candidate extraction unit 35 determines, for each corresponding attribute, between the attribute name of the target attribute and the attribute name of the corresponding attribute. Is calculated (step S47). Note that the similarity between the attribute name of the target attribute and the attribute name of the corresponding attribute is calculated in the same manner as the similarity between the two character string type attributes described above.

  Here, with reference to FIG. 18 and FIG. 19, the similarity between the attribute name of the target attribute and the attribute name of the corresponding attribute calculated by the similar attribute candidate extracting unit 35 will be described in detail. Here, it is assumed that the target similarity list is the above-described similarity list 261 shown in FIG. Further, it is assumed that the target attribute is an “occurrence position” attribute that constitutes the data table 221 of the bank A in the similarity list 261. Note that the threshold value input by the threshold value input unit 34 is 0.8 as described above.

  In this case, in the similarity list 261, attributes (corresponding attributes) whose similarity with the “occurrence position” attribute that is the target attribute is equal to or greater than the threshold (0.8) are “position / occurrence” attributes and “ There are two attributes, “Position / Verifier”. Note that the attribute name of the “occurrence position” attribute is “occurrence position”, the attribute name of the “position / occurrence” attribute is “position / occurrence”, and the attribute name of the “position / verifier” attribute Is a “position / verifier”.

  First, with reference to FIG. 18, the similarity between the attribute name of the “occurrence position” attribute that is the target attribute and the attribute name of the “position / occurrence” attribute that is the corresponding attribute will be described.

  In this case, similarly to the similarity between the two character string type attributes described above, the attribute name of the “occurrence position” attribute that is the target attribute and the attribute name of the “position / occurrence” attribute that is the corresponding attribute are morphologically analyzed. As a result of the processing, a word set of attribute names of the “occurrence position” attribute and a word set of attribute names of the “position / occurrence” attribute are created. The word set created here includes words whose part of speech is a noun, for example.

  Specifically, the word set of the attribute name of the “Gener position” attribute includes the words “Generator” and “Position”. In addition, the word set of the attribute name of the “position / occurrence” attribute includes the words “position” and “occurrence”.

  Here, out of the words included in the word set of the attribute name of the “occurrence position” attribute, the number of words that match the word included in the word set of the attribute name of the “position / occurrence” attribute is 2 (the word “ Accrual ”and“ position ”). Of the words included in the word set of the attribute name of the “position / occurrence” attribute, the number of words matching the word included in the word set of the attribute name of the “position / occurrence” attribute is 2 (the word “ Position ”and“ occurrence ”).

  Further, as described above, since the number of words included in the word set of the attribute name of the “Gener position” attribute is 2, the number of words included in the word set of the attribute name of the “Generator position” attribute is The ratio of the number of words that match the word included in the word set of the attribute name of the “job title / occurrence” attribute is 2/2. Further, since the number of words included in the word set of the attribute name of the “position / occurrence” attribute is 2, the number of words included in the word set of the attribute name of the “position / occurrence” attribute is described above. The ratio of the number of words that match the word included in the word set of the attribute name of the “generator title” attribute is 2/2.

  Accordingly, the similarity between the attribute name of the “occurrence position” attribute that is the target attribute and the attribute name of the “position / occurrence” attribute that is the corresponding attribute is an average value of 2/2 and 2/2, that is, , (2/2 + 2/2) / 2 = 1.

  Next, with reference to FIG. 19, the similarity between the attribute name of the “occurrence position” attribute that is the target attribute and the attribute name of the “position / verifier” attribute that is the corresponding attribute will be described.

  In this case, the attribute name of the “occurrence position” attribute is obtained by performing morphological analysis processing on the attribute name of the “occurrence position” attribute that is the target attribute and the attribute name of the “position / verifier” attribute that is the corresponding attribute. And a word set of attribute names of the “position / verifier” attribute are created. The word set created here includes words whose part of speech is a noun, for example.

  Specifically, as described above, the word set of the attribute name of the “Gener position” attribute includes the words “Generator” and “Position”. Further, the word set of attribute names of the “position / verifier” attribute includes the words “position” and “verifier”.

  Here, of the words included in the word set of the attribute name of the “occurrence position” attribute, the number of words that match the word included in the word set of the attribute name of the “position / verifier” attribute is 1 (the word “ Position ”). Further, among the words included in the word set of the attribute name of the “position / verifier” attribute, the number of words matching the word included in the word set of the attribute name of the “occurrence position” attribute is 1 (the word “position” ]).

  Further, as described above, since the number of words included in the word set of the attribute name of the “Gener position” attribute is 2, the number of words included in the word set of the attribute name of the “Generator position” attribute is The ratio of the number of words that match the word included in the word set of the attribute name of the “position / verifier” attribute is 1/2. In addition, since the number of words included in the word set of the attribute name of the “position / verifier” attribute is 2, the number of words included in the word set of the attribute name of the “position / verifier” attribute is described above. The ratio of the number of words that match the word included in the word set of the attribute name of the “generator position” attribute is ½.

  Accordingly, the similarity between the attribute name of the “occurrence position” attribute that is the target attribute and the attribute name of the “position / verifier” attribute that is the corresponding attribute is an average value of 1/2 and 1/2, that is, , (1/2 + 1/2) / 2 = 1/2.

  Returning to FIG. 17 again, the similar attribute candidate extraction unit 35 determines the corresponding attribute based on the similarity calculated for each corresponding attribute (the similarity between the attribute name of the target attribute and the attribute name of the corresponding attribute). Ranking is performed on the images (step S48). Specifically, the similar attribute candidate extraction unit 35 performs ranking so as to increase the priority of the corresponding attribute having a high degree of similarity. As described above, when the “occurrence position” attribute is the target attribute and the “position / occurrence” attribute and the “position / verifier” attribute are the corresponding attributes, the attribute name of the “occurrence position” attribute and Since the similarity between the attribute name of the “Occupation Position” attribute and the attribute name of the “Position / Occurrence” attribute is higher than the similarity between the attribute names of the “Position / Verifier” attribute, ”Attribute gives higher priority.

  Next, the similar attribute candidate extraction unit 35 extracts similar attribute candidates in step S46. In this case, the similar attribute candidate extraction unit 35 extracts each (combination) of the target attribute and the corresponding attribute as a similar attribute candidate.

  On the other hand, when it is determined in step S44 that the number of corresponding attributes is one, the similar attribute candidate extraction unit 35 extracts similar attribute candidates in step S46. In this case, the similar attribute candidate extraction unit 35 extracts the target attribute and the corresponding attribute (a combination thereof) as similar attribute candidates.

  When it is determined in step S45 that there is no attribute having the maximum similarity with the target attribute, that is, when there is no attribute that satisfies a predetermined condition, or when the process of step S46 is executed, It is determined whether or not the above-described steps S42 to S48 have been executed for all attributes in the similarity list (step S49).

  If it is determined that processing has not been executed for all attributes in the target similarity list (NO in step S49), the process returns to step S42 described above and is repeated. In this case, the process is executed with an attribute for which the process of steps S42 to S48 has not been executed as a target attribute.

  On the other hand, if it is determined that the processing has been executed for all of the target similarity list (YES in step S49), all the values stored in the numerical attribute similarity storage unit 25 and the character string attribute similarity storage unit 26 are stored. It is determined whether or not the above-described steps S41 to S49 have been performed for the similarity list (step S50).

  When it is determined that the process is not executed for all the similarity list (NO in step S50), the process returns to the above step S41 and is repeated. In this case, the process is executed using the similarity list that has not been subjected to the processes of steps S41 to S49 as the target similarity list.

  On the other hand, when it is determined that the process has been executed for all the similarity list (YES in step S50), the similar attribute candidate extraction process is terminated.

  When the similar attribute candidate extraction process is executed as described above, the similar attribute candidates extracted in the similar attribute candidate extraction process are stored in the similar attribute candidate storage unit 27.

  Here, FIG. 20 shows an example of the data structure of the similar attribute candidate storage unit 27. The similar attribute candidate storage unit 27 stores two attributes (attribute names) constituting different data tables extracted as similar attribute candidates as described above in association with each other.

  In the example illustrated in FIG. 20, the similar attribute candidate storage unit 27 includes, for example, an “occurrence cause” that configures the data table 222 of the B bank in association with the “cause / occurrence” attribute that configures the data table 221 of the A bank Attribute is stored. The similar attribute candidate storage unit 27 includes, for example, an “occurrence cause / occurrence” attribute that configures the data table 223 of the C bank in association with the “occurrence cause / occurrence” attribute that configures the data table 221 of the A bank. Is stored.

  According to this, the “occurrence / occurrence” attribute constituting the bank A data table 221 and the “occurrence cause” attribute constituting the bank B data table 222 are attribute candidates representing the same contents (that is, similar Attribute candidate). Similarly, the “occurrence cause / occurrence” attribute constituting the bank A data table 221 and the “occurrence cause / occurrence” attribute constituting the bank C data table 223 are candidates for attributes representing the same contents. It is shown.

  Further, the similar attribute candidate storage unit 27 includes, for example, a “position / occurrence” attribute and a “position” that configure the data table 222 of the B bank in association with the “generation position” attribute that configures the data table 221 of the A bank. / Verifier attribute is stored. According to this, the “generator position” attribute constituting the bank A data table 221 and the “position / generator” attribute constituting the bank B data table 222 are candidate attributes representing the same content. It is shown. In addition, it is shown that the “generator position” attribute that configures the data table 221 of the A bank and the “position / verifier” attribute that configures the data table 222 of the B bank are candidates for attributes that express the same content. Yes.

  Note that the “position / occurrence” attribute and the “position / verifier” stored in the similar attribute candidate storage unit 27 in association with the “occurrence position” attribute constituting the data table 221 of the bank A are as described above. In the similar attribute candidate extraction process, they are arranged in the order of priority assigned to the “position / occurrence” attribute and the “position / verifier” attribute (corresponding attribute).

  As shown in FIG. 20, the similar attribute candidate storage unit 27 stores all combinations of two attributes extracted as similar attribute candidates in the similar attribute candidate extraction process.

  The similar attribute candidates stored in the similar attribute candidate storage unit 27 shown in FIG. 20 are held in, for example, the data tables 221 to 223 (banks A to C) stored in the data table storage unit 22 as described above. Data) can be used when comparing the data tables 221 to 223 of the banks A to C in the data analysis.

  As described above, in the present embodiment, the words constituting the character strings included in the attribute values of the character string type attributes constituting the plurality of tables stored in the table storage unit 22 are extracted, and the extracted By calculating the similarity between two character string type attributes constituting different tables based on the word, and extracting the two character string type attributes as similar attribute candidates based on the calculated similarity degree, the attribute Appropriate attributes can be associated between arbitrary data tables by using only the characteristics of the data and considering the meaning of the attribute value.

  Further, in the present embodiment, two numerical type attributes that configure different tables based on the range of numerical values included in the attribute values of the numerical type attributes that configure the plurality of tables stored in the table storage unit 22. By calculating the similarity between the two, and extracting the two numeric type attributes as similar attribute candidates based on the calculated similarity, the numeric type attributes are also appropriately associated between arbitrary data tables. be able to.

  Furthermore, in this embodiment, by classifying the numerical type attributes constituting a plurality of tables into numerical type attribute subcategories, for example, the similarity between two numerical type attributes that cannot be compared in numerical range is calculated. This can be avoided and appropriate association of numeric type attributes can be performed.

(Second Embodiment)
Next, a second embodiment will be described. The hardware configuration and functional configuration of the data analysis support apparatus according to this embodiment are the same as those of the first embodiment described above, and will be described with reference to FIGS. 1 and 2 as appropriate.

  In the present embodiment, the point that the character string type attribute processing unit 33 included in the data analysis support device 30 uses the synonym dictionary when calculating the similarity between the two character string type attributes is the first described above. Different from the embodiment.

  Here, FIG. 21 is a block diagram illustrating a functional configuration of the character string type attribute processing unit 33 included in the data analysis support device 30 according to the present embodiment.

  The character string type attribute processing unit 33 includes a synonym dictionary storage unit 334 and a character string type attribute similarity calculation unit 335. In the present embodiment, the synonym dictionary storage unit 334 is stored, for example, in the external storage device 20 shown in FIG.

  The synonym dictionary storage unit 334 stores in advance a synonym dictionary in which a plurality of words having similar meanings (that is, having similar meanings) are registered as synonyms.

  The character string type attribute similarity calculation unit 335 generates a different data table based on the word set of each character string type attribute stored by the attribute value word set storage unit 331 and the synonym dictionary stored in the synonym dictionary storage unit 334. The degree of similarity between the two character string type attributes that constitutes is calculated. In this case, the character string type attribute similarity calculation unit 335 has two character string type attributes (first and second character strings) in the same manner as the character string type attribute similarity calculation unit 333 in the first embodiment described above. The number of words that match between word sets of type attribute) is specified. At this time, the words that match between the word sets of the two character string type attributes specified by the character string type attribute similarity calculation unit 335 are stored in the synonym dictionary storage unit 334 as well as the completely matched words. Words having similar meanings registered in the synonym dictionary. That is, in the character string type attribute similarity calculation unit 335 in the present embodiment, words having similar meanings are also considered to match.

  The character string type attribute similarity calculation unit 335 has the same function as the character string type attribute similarity calculation unit 333 in the first embodiment except for this point.

  Here, the operation of the data analysis support device 30 according to the present embodiment will be described. Of the processes executed in the data analysis support device 30 according to the present embodiment, processes other than the character string type attribute similarity calculation process included in the similarity calculation process (the process of step S4 shown in FIG. 8 described above). Since is the same as that of the first embodiment described above, detailed description thereof is omitted.

  Hereinafter, the processing procedure of the similarity calculation processing of the character string type attribute in this embodiment will be described. Here, for convenience, description will be made with reference to the flowchart of FIG.

  First, the processes of steps S21 to S23 shown in FIG. 13 are executed. Since the processes in steps S21 to S23 are the same as those described in the first embodiment, detailed description thereof is omitted.

  When it is determined in step S23 that processing has been executed for all the character string type attributes stored in the character string type attribute storage unit 24, the character string type attribute similarity calculation unit 335, for example, the character string type attribute storage unit 24 The processing of the following steps S24 to S26 is executed for each of the combinations of two character string type attributes (character string type attribute pairs) constituting different data tables stored in. Here, the character string type attribute pair that is the target of this processing is referred to as a target character string type attribute pair. One character string type attribute included in the target character string type attribute pair is referred to as a first character string type attribute, and the other character string type attribute is referred to as a second character string type attribute.

  In this case, the character string type attribute similarity calculation unit 335 acquires the word set of the first and second character string type attributes included in the target character string type attribute pair from the attribute value word set storage unit 332.

  The character string type attribute similarity calculating unit 335 refers to the acquired word sets of the first and second character string type attributes and the synonym dictionary stored in the synonym dictionary storage unit 334, and performs the first and second The number of words that match between the word sets having the character string type attribute of 2 is specified (step S24). In this case, the character string type attribute similarity calculation unit 335 includes a word that matches a word included in the word set of the second character string type attribute and a word that matches the word included in the word set of the second character string type attribute. The number of similar words (the number of matches of the first character string type attribute) is specified. In addition, the character string type attribute similarity calculation unit 335 includes words and similarities that match words included in the word set of the second character string type attribute among words included in the word set of the second character string type attribute. The number of words to be identified (the number of matches of the second character string type attribute) is specified.

  The character string type attribute similarity calculating unit 335 calculates the degree of similarity between the first and second character string type attributes based on the number of matches between the specified first and second character string type attributes ( Step S25). In this case, the character string type attribute similarity calculation unit 335 performs matching of the words included in the word set of the first character string type attribute (first character string type attribute), as in the first embodiment described above. And the matching rate of words included in the word set of the second character string type attribute (word matching rate of the second character string type attribute) are used to calculate the similarity.

  Here, the similarity between the first and second character string type attributes calculated by the character string type attribute similarity calculating unit 335 will be specifically described with reference to FIG.

  Here, the first character string type attribute is the “cause / occurrence” attribute that constitutes the bank A data table 221 shown in FIG. 5, and the second character string type attribute is shown in FIG. It is assumed that it is an “occurrence cause” attribute constituting the data table 222 of the B bank shown.

  As shown in FIG. 22, the word set of the first character string type attribute (that is, the “occurrence cause / occurrence” attribute) has 11 words, specifically, the words “experience” and “insufficient”. , “Guidance”, “education”, “third party”, “accident”, “customer”, “request”, “miss”, “entry” and “error”. The word set of the second character string type attribute (that is, the “occurrence cause” attribute) has 10 words, specifically, the words “knowledge”, “experience”, “education”, “insufficient”. , “Customer”, “accident”, “complexity”, “work”, “content” and “careless mistake”.

  It is assumed that the words “teaching” and “education” are registered as synonyms in the synonym dictionary stored in the synonym dictionary storage unit 334 included in the character string type attribute processing unit 33. It is assumed that the words “customer” and “customer” are registered as synonyms in the synonym dictionary. Further, it is assumed that the words “miss”, “error”, and “careless miss” are registered as synonyms in the synonym dictionary.

  Here, among words included in the word set of the first character string type attribute, words that completely match the word included in the word set of the second character string type attribute are the words “experience” and “insufficient”. , "Education" and "Accident". Further, according to the above synonym dictionary, among words included in the word set of the first character string type attribute, words having meanings similar to words included in the word set of the second character string type attribute (that is, Synonyms) are a word “teaching” having a similar meaning to the word “education” included in the word set of the second character string type attribute, a word “customer” having a similar meaning to the word “customer”, and a word “careless mistake”. The words “Miss” and “Miss” are similar in meaning. In this case, the number of matches of the first character string type attribute is 8 (words “experience”, “insufficient”, “teaching”, “education”, “accident”, “customer”, “miss” and “error”) Become.

  On the other hand, among words included in the word set of the second character string type attribute, words that completely match the words included in the word set of the first character string type attribute are the words “experience”, “education”, “Insufficient” and “Accident”. Further, according to the above synonym dictionary, among words included in the word set of the second character string type attribute, words having meanings similar to words included in the word set of the first character string type attribute (that is, Synonyms) include a word “education” similar in meaning to the word “guidance” included in the word set of the first string attribute, a word “customer” similar in meaning to the word “customer”, and a word “miss” "And" Error "are words" Careless Miss "that have similar meanings. In this case, the number of matches of the second character string type attribute is 6 (the words “experience”, “education”, “insufficient”, “customer”, “accident”, and “careless mistake”). For example, when there are a plurality of words and similar words that completely match a word included in the word set of the first character string type attribute, such as the words “education” and “careless mistake”, the same word Are treated as one.

  In FIG. 22, two completely matching words are connected by a solid line, and two words having similar meanings are connected by a broken line.

  As described above, since the number of words included in the word set of the first character string type attribute is 11, the word matching rate of the first character string type attribute is 8/11. Further, since the number of words included in the word set of the second character string type attribute is 10, the word matching rate of the second character string type attribute is 6/10.

  Accordingly, the similarity between the first and second character string type attributes is calculated as an average value of 8/11 and 6/10, that is, (8/11 + 6/10) /2≈0.664. .

  Returning to FIG. 13 again, the processing of steps S25 to S27 is executed. Note that the processing in steps S25 to S27 is the same as that described in the first embodiment, and a detailed description thereof will be omitted.

  As described above, in the present embodiment, the character string type attribute specified when the similarity between two character string type attributes is calculated by using the synonym dictionary stored in the synonym dictionary storage unit 334. Even if a word that is similar in meaning (synonyms) is included as a word that matches between the word sets in the first embodiment, the word is not treated as a word that matches between the word sets in the first embodiment. If they are similar words, they are treated in the same way as the matching words, and therefore the similarity between the two character string type attributes can be calculated more appropriately.

  In the present embodiment, the synonym dictionary is used when calculating the similarity between two character string type attributes (specifying the number of matching words between word sets). The synonym dictionary may be used when calculating the similarity between the attribute names described in the first embodiment.

  According to at least one embodiment described above, it is possible to provide a data analysis support apparatus and program capable of associating appropriate attributes between arbitrary data tables.

  Note that the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in each embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

  DESCRIPTION OF SYMBOLS 10 ... Computer, 20 ... External storage device, 22 ... Data table storage part, 23 ... Numerical value attribute storage part, 24 ... Character string type attribute storage part, 25 ... Numerical value type attribute similarity storage part, 26 ... Character string type attribute Similarity degree storage unit, 27 ... similar attribute candidate storage unit, 31 ... attribute type classification unit, 32 ... numeric type attribute processing unit, 33 ... character string type attribute processing unit, 34 ... threshold input unit, 35 ... similar attribute candidate extraction unit , 321 ... Attribute value numerical range specifying unit, 322 ... Attribute value numerical range storage unit, 323 ... Numeric attribute similarity calculation unit, 331 ... Attribute value word extraction unit, 332 ... Attribute value word set storage unit, 333 ... Character string Type attribute similarity calculation unit, 334... Synonym dictionary storage unit, 335... Character string type attribute similarity calculation unit.

Claims (6)

A first data table composed of a first attribute including a first character string type attribute having an attribute value including a character string and a second including a second character string type attribute having an attribute value including a character string Data table storage means for storing in advance a second data table comprising the attributes of:
Extracting the first word constituting the character string included in the attribute value of the first character string type attribute included in the first attribute constituting the first data table stored in the data table storage means First word extracting means for
Extracting the second word constituting the character string included in the attribute value of the second character string type attribute included in the second attribute included in the second data table stored in the data table storage means Second word extracting means for
Based on the first word extracted by the first extraction means and the second word extracted by the second extraction means, the first attribute included in the first data table is included in the first attribute included in the first data table. Similarity calculating means for calculating the similarity between the first character string type attribute and the second character string type attribute included in the second attribute constituting the second data table;
Based on the calculated similarity, the first character string type attribute included in the first attribute constituting the first data table and the second attribute constituting the second data table are included. A data analysis support apparatus comprising: similar attribute candidate extraction means for extracting the second character string type attribute as a similar attribute candidate.   The similarity calculation means matches the ratio of the number of the first words that matches the second word with respect to the number of the first words, and matches the first word with respect to the number of the second words The data analysis support according to claim 1, wherein the similarity between the first character string type attribute and the second character string type attribute is calculated based on a ratio of the number of the second words. apparatus. The first attribute constituting the first data table stored in the data table storage means includes a plurality of first character string type attributes,
The second attribute constituting the second data table stored in the data table storage means includes a plurality of second character string type attributes,
The first word extraction unit is configured to output the first character string for each first character string type attribute included in the first attribute constituting the first data table stored in the data table storage unit. Extracting a first word constituting a character string included in an attribute value of a type attribute;
The second word extraction unit is configured to output the second character string for each second character string type attribute included in the second attribute that constitutes the second data table stored in the data table storage unit. Extracting a second word constituting the character string included in the attribute value of the type attribute;
The similarity calculation means includes attributes of the first character string type attribute extracted by the first extraction means for each combination of the first character string type attribute and the second character string type attribute. Based on the first word constituting the character string included in the value and the second word constituting the character string included in the attribute value included in the second character string type attribute extracted by the second extracting means Calculating the similarity between the first string type attribute and the second string type attribute,
The similar attribute candidate extraction unit is configured to generate another first character string type attribute and a second character string based on the similarity calculated for each combination of the first character string type attribute and the second character string type attribute. The first character string type attribute and the second character string type attribute that have a maximum degree of similarity with each other as compared with the degree of similarity with the character string type attribute are extracted as similar attribute candidates. The data analysis support device described. The first attribute constituting the first data table stored in the data table storage means further includes a first numeric type attribute having an attribute value including a numeric value,
The second attribute constituting the second data table stored in the data table storage means further includes a second numeric type attribute having an attribute value including a numeric value,
The similarity calculation means includes:
A numerical value range included in an attribute value included in the first numerical value attribute included in the first attribute and a numerical value range included in an attribute value included in the second numerical value attribute included in the second attribute. Based on this, the similarity between the first numeric type attribute and the second numeric type attribute is calculated,
The similar attribute candidate extracting means calculates the first numerical value attribute and the second numerical value attribute based on the calculated similarity between the first numerical value attribute and the second numerical value attribute. The data analysis support apparatus according to claim 1, wherein the data analysis support apparatus is extracted as a similar attribute candidate. First classification means for classifying the first numeric type attribute according to a numeric value included in an attribute value of the first numeric type attribute;
A second classification means for classifying the second numeric type attribute according to a numeric value included in an attribute value of the second numeric type attribute;
The similarity calculation means calculates the similarity between the first numeric type attribute and the second numeric type attribute having the same classification destination by the first classification means and the second classification means. 5. The data analysis support apparatus according to claim 4, wherein A first data table composed of a first attribute including a first character string type attribute having an attribute value including a character string and a second including a second character string type attribute having an attribute value including a character string In a data analysis support apparatus composed of an external storage device having a data table storage means for preliminarily storing a second data table composed of the attributes of and a computer using the external storage device, the data analysis support device is executed by the computer A program
In the computer,
Extracting the first word constituting the character string included in the attribute value of the first character string type attribute included in the first attribute constituting the first data table stored in the data table storage means And steps to
Extracting the second word constituting the character string included in the attribute value of the second character string type attribute included in the second attribute included in the second data table stored in the data table storage means And steps to
Based on the extracted first word and the extracted second word, the first character string type attribute and the second data included in the first attribute constituting the first data table Calculating the similarity of the second character string type attribute included in the second attribute constituting the table;
Based on the calculated similarity, the first character string type attribute included in the first attribute constituting the first data table and the second attribute constituting the second data table are included. Extracting the second character string type attribute as a similarity attribute candidate.

Images