JP2010218386A - Method of matching data between business models and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform data matching between business models each composed of a data model and a process model. <P>SOLUTION: A data model and a process model corresponding to a first business model are integrated into one graph based on a relation between elements of the models to generate a first extension process graph on data. Similarly, a data model and a process model corresponding to a second business model are integrated into one graph based on a relation between elements of the models to generate a second extension process graph on data. Thereafter, the elements of the first extension process graph and the nodes of the second extension process graph are paired, and a pair-wise connection graph describing a relation each between pairs by labeling is generated on data. A similarity propagation graph in which similarity and propagation coefficient between pairs are set is generated on data, and the similarity each between pairs is calculated by repetitive operation. Thereafter, a pair with high similarity obtained by filtering is presented on a display device. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a signal processing technique for data matching of a plurality of business models. For example, the present invention relates to a technique for data matching between a business model used in a certain company and a business model used as an industry reference through signal processing.

  Currently, industry groups are working to standardize business models for the purpose of reducing investment in IT infrastructure and interconnecting them. A business model to be a reference developed by an industry group is referred to as a “reference business model” in this specification. Reference service models include, for example, NGOSS (New Generation Operation Systems and Software) in the telecom field, HL7 (Healthcare Level Seven) in the health care field, and ACORD (Association for Cooperative Operations Research and Development).

  However, when the reference business model is deployed in a company, matching between the reference business model and an existing business model in the company becomes a problem. In general, a business model is composed of a data model representing data to be exchanged, and a process model representing processing and its procedure. For this reason, when the two models are combined, the number of items to be matched reaches tens of thousands in some cases. However, at present, a technology for matching both business models has not been established. For this reason, the current matching work is performed manually, and a large man-hour is a problem.

  Currently, matching between metamodels (schema) in a data model is known as a technique for automating matching of a plurality of business models. FIG. 1 shows a specific example of the meta model 101 (FIG. 1A) of the reference data model corresponding to the order slip and a specific example of the meta model 102 (FIG. 1B) of the corresponding existing data model. The data model is generally given as a UML class diagram or a data flow diagram. In the following description, as shown in FIG. 1, the data model is represented by this UML notation. As described above, in existing methods, matching is generally performed at the metamodel level. In the existing matching technology, the class / attribute name / type similarity and the relationship between classes such as inheritance / aggregation / dependency are used for matching.

JP 2007-188343 A

Melnik, `` Generic Model Management '', Springer 2004, p.117-131

  However, a mechanism for integrating the entire business model composed of the data model and the process model by data matching has not yet been realized. In particular, in the case of a process model, the above-described matching method at the metamodel level for the data model cannot be applied as it is. This is because process models created based on a common metamodel are subject to matching.

  FIG. 2 shows a specific example of the reference process model 210 (FIG. 2B) for a purchasing operation and a specific example of the corresponding existing process model 220 (FIG. 2C). Here, the process model is represented by a graph such as a BPMN (Business Process Management Notation) or a UML activity diagram. In the following description, a process model is expressed by this BPMN notation. A metamodel 201 of the process model is shown in FIG. The meta model 201 is represented by a node (Node) and an edge (Edge) representing a graph. As described above, both the reference process model 210 and the existing process model 220 are created based on the meta model 201.

  Accordingly, the process model matching level needs to be executed at a level different from the data model matching level (ie, meta model). In addition, in the matching at the process model level, the ambiguity peculiar to the process model, in which the description is structurally different even though the process model is semantically the same, becomes a problem.

  Therefore, the inventor has the following mechanism that can handle the data model and the process model equally and can solve the ambiguity of the process model in order to realize the data matching process at the business model level. provide.

  First, the data processing unit generates a first extended process graph on the data by integrating them into one graph based on the relationship between the data model corresponding to the first business model and the elements of the process model. . Similarly, the data processing unit generates a second extended process graph on the data by integrating them into one graph based on the relationship between the data model corresponding to the second business model and the elements of the process model. To do. Thereafter, the data processing unit forms a pair with the elements of the first extended process graph and the elements of the second extended process graph, and creates a pair-wise connection graph describing the relationship between each pair with a label on the data. Generate. Next, the data processing unit generates a similarity propagation graph in which the similarity between the pairs and the propagation coefficient are set on the data, and calculates the similarity between the pairs by repeated calculation. Thereafter, the data processing unit presents a pair having high similarity on the display device by filtering.

  According to the present invention, the number of man-hours required for matching can be greatly reduced as compared with the conventional method. In addition, if the extended process graph proposed by the inventor is used, the data model and the process model can be handled in a unified manner. As a result, the entire two business models can be data-matched. Thereby, matching accuracy can be improved.

It is a figure explaining the specific example of a data model. It is a figure explaining the specific example of a process model. It is a figure which shows the outline | summary of the system configuration | structure of the matching system which concerns on an example. It is a flowchart which shows the outline | summary of the matching process which concerns on an example. It is a flowchart which shows the example of an execution procedure of the matching process which concerns on an example. It is a figure which shows the normalization operation example of a process model. It is a figure explaining the specific example of an extended process graph. It is a figure explaining the specific example of a pairwise connection graph. It is a figure explaining the specific example of a similarity propagation graph.

  An example of a matching system that executes data matching processing between business models will be described below.

(1) Configuration Example (1-1) Configuration of Matching System FIG. 3 shows a configuration example of a matching system that executes matching processing between business models. The matching system according to the embodiment includes a computer 300, an input device 320, and a display device 330. The computer 300 includes a CPU 301 that executes data computation, a ROM 302, a RAM 303, a hard disk drive 306, a CPU bus 312 that realizes data transfer between these devices, and interfaces 309 to 311 that couple these devices to the CPU bus 312. The

  Incidentally, the RAM 303 has at least an execution area for the inter-business-model matching program 304 for causing the CPU 301 to execute arithmetic processing, and a work area 305 for storing data temporarily generated at the time of arithmetic operation. In addition, in the storage area of the hard disk drive 306, at least a program storage unit 307 as a storage area for the business model matching program 304 and a data storage unit 308 as a data storage area for the reference business model and the existing business model are secured. Is done.

  In the description of the embodiment, it is assumed that the matching process is realized as one function of a program executed on the computer 300, but the matching process may be realized as an apparatus dedicated to the matching process. In this case, the processing function can be realized as an ASIC or a dedicated processing board. Incidentally, these hardware configurations correspond to the “data processing unit” in the claims.

  Further, the business model matching program 304 can be written to the hard disk drive 306 from a recording medium in which the program is recorded, or through a broadcast signal or a communication signal including the program.

  In this embodiment, the business model matching program 304 is written in the hard disk drive 306, but the recording medium is read by a flexible disk, CD-ROM, DVD-ROM, semiconductor memory or other computer. Any possible storage medium may be used.

(1-2) Business Model Matching Operation (a) Schematic Operation FIG. 4 shows an overview of a data matching process executed through the business model matching program 304. This data matching process is started by an operation input by the user through the input device 320 (step 401). When the computer 300 detects the operation input, the computer 300 reads, from the data storage unit 308 of the hard disk drive 306, the data of the reference business model and the data of the existing business model that are designated and input as matching targets. Here, the reference business model data includes a meta model 101 and a reference process model 210 corresponding to the reference data model. The existing business model data includes the meta model 102 and the existing process model 220 corresponding to the existing data model. The computer 300 stores the read data in the work area 305 of the RAM 303 (step 402).

  The computer 300 executes a matching operation between the business model data stored in the work area 305 in accordance with the processing procedure described in the business model matching program 304 (step 403). Details of the processing procedure will be described later. The calculation result obtained by the matching operation of the computer 300 is stored in the work area 305.

  The computer 300 displays the calculation result stored in the work area 305 on the display device 330 (step 404). As a result, the processing operation as the business model matching program 304 ends (step 405).

  On the other hand, the user who has received the calculation result through the display device 330 starts verification work on the matching result that is likely to be presented. The verification work here may be performed on a highly accurate matching result. Therefore, the user's work man-hours can be greatly reduced.

(B) Detailed Operation Performed in Matching Calculation Next, the detailed operation of the matching calculation performed in Step 403 will be described. FIG. 5 shows an outline of a processing procedure executed in the matching calculation.

(Step 502)
After the start of the matching calculation, the computer 300 executes an initial matching process. In this initial matching process, the computer 300 compares the meta model 101 of the reference data model and the meta model 102 of the existing data model by a character string for each element, and compares the reference process model 210 and the existing process model 220 for each element. And a matching process for comparing with a character string. For example, in the case of the example in FIG. 2, the process “verification” of the reference process model 210 and the process “verification” of the existing process model 220 have the same name. In this case, the computer 300 determines that these elements match.

(Step 503)
Next, the computer 300 refers to the reference process model 210 and the existing process model 220, searches for a structure that can have a plurality of structures with the same meaning, and unifies the structure into a predetermined specific structure. Execute. That is, a process model normalization process is executed. FIG. 6 shows an example of normalization operation of the process model. FIG. 6A shows an example in which a process model 600 having a return side is rewritten to a process model 601 for loop processing. 6B is an example in which a process model 610 having continuous processing that does not exchange data with each other is rewritten to a process model 611 that executes parallel processing.

  Thus, by performing normalization processing on the process model in advance, the difference in expression on the model structure can be changed to a unified expression. Thereby, matching accuracy can be improved. Also, by eliminating the return edge and determining the processing order during normalization, it is possible to make correspondence between elements and labeling edges on the extended process graph described later.

(Step 504)
Thereafter, the computer 300 generates a graph in which the data model is integrated with the process model (hereinafter referred to as “extended process graph”) for each business model (step 504). That is, the computer 300 creates an extended process graph 710 (FIG. 7A) corresponding to the reference business model based on the relationship between the elements of the reference data model metadata 101 and the reference process model 210. Similarly, based on the relationship between the metadata 102 of the existing data model and the elements of the existing process model 220, an extended process graph 720 (FIG. 7B) corresponding to the existing business model is created.

Here, the extended process graph G _ep is an integration of the process model and the data model, and is defined as shown in the following equation (1).

Incidentally, the extended process graph G _ep is defined by a set of nodes V _ep and a set of edges E _ep . The clause set V _ep is defined as a combination of the process model clause set V ^process and the data model clause set V ^data . However, as shown in FIG. 7, when one or more virtual clauses are defined on the edge between the process model and the data model, the set of these clauses is also included in the clause set V _ep. . The edge set E _ep is defined as a combination of a process model edge set E ^process , a data model edge set E ^data, and an edge set E ^cross between the process model and the data model. In this specification, in both the process model and the data model, the edge (v _s , l, v _t ) is the start clause v _s. And the end clause v _t And a combination with the label l.

In the extended process graph G _ep in this embodiment, the data model is distinguished by attaching different labels to the edge derived from the data model, the edge derived from the process model, and the edge connecting the data model and the process model. To do. For example, in the case of the extended process graphs 710 and 720, the side derived from the process model is labeled L1 or L2, the side derived from the data model is labeled O1, and the side connecting the process model and the data model Is labeled C1. Labels “attribute” and “type” representing the relationship between elements are attached to the other sides as labels.

Further, in the extended process graph _Gep in this embodiment, the elements matched by the initial matching in step 502 are distinguished in terms of data by attaching different labels on the upstream side and the downstream side in terms of the process. In the case of this embodiment, in step 502, the process “verification” of the reference process model 210 matches the process “verification” of the existing process model 220. For this reason, in the extended process graph 710 shown in FIG. 7A, the label L1 is attached to the upstream side of the element RT6 corresponding to the process “verification”, and the label L2 is attached to the downstream side of the element RT6. It is attached. Similarly, in the extended process graph 720 shown in FIG. 7B, the label L1 is attached to the upstream side of the element AT5 corresponding to the process “verification”, and the label L2 is attached to the downstream side of the element AT5. It is attached. As a result, it is possible to prevent the elements after the element RT6 of the extended process graph 710 corresponding to the reference work model from matching with the elements before the element AT5 of the extended process graph 720 corresponding to the existing work model.

Further, in the extended process graph _Gep in this embodiment, a different label is also provided between the element of the node derived from the data model and the element of the node derived from the process model to distinguish on the data. Moreover, distinguishing the data given different labels and expansion process graph G _ep corresponding to the existing business model and expansion process graph G _ep corresponding to the reference work model. In addition, the elements of the sections are serially numbered from the upstream side in the model with the same origin to distinguish them in the data. As a result, the positional relationship of the elements of the extended process graph 710 and the extended process graph 720 in the process is clarified, and when the pairwise connection graph described later is generated, It is possible to create pairs of elements without excess or deficiency.

(Step 505)
Thereafter, the computer 300 creates a pair-wise connection graph that represents the connection relationships of all the possible pairs for each element of the extended process graph 710 and each element of the extended process graph 720 generated in step 504 in a graph. FIG. 8 shows a part of a pair-wise connection graph created from the extended process graph of FIG. In FIG. 8, a pair-wise connection graph is created by pairing elements having the same label among the nodes constituting the extended process graph. Incidentally, the subgraphs 810 (FIG. 8A), 820 (FIG. 8B), 830 (FIG. 8C), and 840 (FIG. 8D) have side labels L1, L2, It is an example of activityType, C1.

The pairwise connection graph G _pc is a graph for viewing the relationship between the clause of the extended process graph G _{ep (reff)} corresponding to the reference business model and the clause of the extended process graph G _{ep (AsIs)} corresponding to the existing business model. Yes, it is defined as shown in Equation (2) below. The elements of the clause set V _pc are the elements of the extended process graph clause V _{ep (reff)} corresponding to the reference business model and the elements of the extended process graph clause set V _{ep (AsIs)} corresponding to the existing business model. Defined as a pair. The edge set E _pc has the same label l among the edge set E _{pc (reff)} of the extended process graph corresponding to the reference business model and the edge set E _{pc (AsIs)} of the extended process graph corresponding to the existing business model. Consists of sides.

  At this time, in order to reduce the amount of calculation, pairs of clauses are created in consideration of the execution order of the normalized process model. In the case of this embodiment, number i is assigned in order to the nodes derived from the process model in the extended process graph 710, and order j is assigned to the nodes derived from the process model in the extended process graph 720. Yes.

Here, a clause v _{i0 (reff)} of the extended process graph 710 corresponding to the reference business model and a clause v _{j0 (AsIs) of} the extended process graph corresponding to the existing business model are paired, and a new clause of the pair-wise connection graph. Consider the case of generating (v _{i0 (reff)} , v _{j0 (AsIs)} ). At this time, the node v _{i0 + 1 (reff)} of the next extended process graph 710 forms a pair only with the node after the node v _{j0 + 1 (AsIs)} of the extended process graph 720. Thus, for example, after creating the pair (RT2, AT2) from the extended process graph of FIG. 7, it is possible to prevent the creation of a pair (RE1, AT3) in which the execution order of the original process model is reversed, and the calculation amount Can be reduced.

(Step 506)
Thereafter, the computer 300 creates a similarity propagation graph based on the pair-wise connection graph created in step 505 (FIG. 8). FIG. 9 shows similarity propagation graphs 910 (FIG. 9A), 920 (FIG. 9B), 930 (FIG. 9C), and 940 corresponding to the pairwise connection graphs 810, 820, 830, and 840, respectively. (FIG. 9D) is shown. In particular, the similarity propagation graph 940 is a similarity propagation graph that connects a process model and a data model. The existence of the similarity propagation graph 940 ensures mutual propagation of similarity between a pair of clauses between process models and a pair of clauses between data models.

The affinity propagation graph G _sp, pairwise connection graph G _pc sections _{(v reff, v AsIs) ∈V} pc similarity between pairs v _reff and v _AsIs with the paragraph this similarity is how adjacent propagated It is a graph expressing whether or not to be defined, and is defined by the following formula (3). As shown in the equation (3), the similarity propagation graph G _sp includes a node set V _sp , an edge set E _sp , a similarity function σ, a similarity set Σ, a propagation function ω, and a propagation coefficient set Ω. Defined. Note that the edge of the pair-wise connection graph G _pc is unidirectional, whereas the edge set E _sp in the similarity propagation graph G _sp is composed of bidirectional edges.

  Here, the value of each element of the pair similarity set Σ is assumed to be a real value from 0 to 1. There are a plurality of ways of giving initial values for the similarity Σ. For example, when all the initial values are set to 1, when both pairs are texts, the similarity between the texts is obtained from the thesaurus or the edit distance, A method of using it as an initial value of similarity is used. Similarly, the value of each element of the set of propagation coefficients Ω is given as a real value from 0 to 1. Note that the initial value is given so that the sum of the propagation coefficients for all sides where the node of interest is the start node is 1. There are multiple ways to assign initial values to each side. For example, a method in which the target node is the same for all the sides that are the starting clause, or the side is derived from the process model or data Use a method that gives different values depending on the edges from the model.

(Step 507)
Next, the computer 300 calculates the similarity of each pair in the created similarity propagation graph (for example, 910 to 940) based on the following formula (4), and repeats the calculation process until the calculation result converges: Execute. That is, the computer 300 repeatedly performs a calculation process until the difference between the value of the similarity function σ ^{k + 1 and} the value of σ ^k for a given pair is equal to or less than a given threshold value. Here, k is the number of repetitions. Note that the second term of Equation (4) is the similar component value between the node to be calculated and the node located downstream of the process, and the third term is the process for the node to be calculated. It is a similar component value between the nodes located on the upstream side.

(Step 508)
Thereafter, the computer 300 compares the similarity between the pairs calculated in the previous step 507 with a threshold value, and executes a filtering process for excluding the pairs having a similarity smaller than the threshold value. The threshold value is corrected in a timely manner in consideration of the number of work steps by the user, the number of pairs remaining as a filtering result, and the like. Incidentally, the input device 320 is used to correct the threshold value.

(1-3) Summary As described above, the process model is initialized by initializing the notation variation appearing in each process model to be subjected to the matching process to the notation in which the ambiguity of the return side and the order relation is excluded. The order relationship can be clarified.

  In addition, an extended process graph that represents the model of the process model that has been initialized and the meta model of the data model is created on the same graph. Label it. Similarly, in the initial matching, each side is given a different label based on whether it is located before or after the process for the element of the clause whose matching is confirmed in the comparison between the process models. This makes it possible to clarify the matching range when creating a pair-wise connection graph for the extended process graph in consideration of the processing order on the process. As a result, it is possible to avoid unnecessarily widening the pair search range by the computer 300. The amount of calculation can be reduced accordingly.

  In the case of this form example, when creating the similarity propagation graph, the similarity between texts is obtained using a thesaurus or editing distance for the clauses that are pairs of text, and this is used as the initial value of similarity between pairs. By giving, the reliability with respect to the calculated similarity between pairs can be improved. In addition, when creating a similarity propagation graph, the reliability of similarity between calculated pairs can be improved by changing the initial value of the propagation coefficient depending on whether the edge is derived from a model with high reliability. . In particular, if there is a difference in reliability between the process model and the data model that make up the business model to be matched, it is finally calculated by increasing the influence of the propagation coefficient from the higher reliability. The reliability of the similarity between each pair can be increased.

(2) Other Embodiments In the case of the above-described embodiment examples, the case has been described in which business models are subjected to data matching using a purchase operation as a specific example. However, it goes without saying that the business model to be matched is not limited to this.

  In the description of the example, for convenience, one business model (data model, process model) to be matched is called a reference model, and the other business model (data model, process model) is called an existing model. One of the targets need not be a reference business model formulated by an industry group.

  In the case of the above-described embodiment, the case where the process model normalization process (step 503) is executed after the initial matching (step 502) has been described. However, the execution order may be changed.

300 ... Computer 301 ... CPU
302 ... ROM
303 ... RAM
306 ... HD
320 ... Input device 330 ... Display device

Claims (9)

A process in which the data processing unit reads the data of the data model and the process model corresponding to the first business model and the data model and the process model of the data corresponding to the second business model from the storage device;
The data processing unit compares the character string of each element constituting the process model corresponding to the first business model and the character string of each element constituting the process model corresponding to the second business model on the data And a process for detecting an element with a matching relationship,
A process in which the data processing unit normalizes the model notation of the process model corresponding to the first and second business models into a notation using a preset model notation;
A process in which the data processing unit generates a first extended process graph on data by integrating these into one graph based on the relationship between the data model corresponding to the first business model and the elements of the process model When,
A process in which the data processing unit generates a second extended process graph on data by integrating these into one graph based on the relationship between the data model corresponding to the second business model and the elements of the process model When,
The data processing unit forms a pair with the elements of the first extended process graph and the elements of the second extended process graph and generates a pair-wise connection graph describing the relationship between each pair with a label on the data Processing to
A process in which the data processing unit generates on the data a similarity propagation graph in which the similarity and propagation coefficient between pairs are set;
A process in which the data processing unit repeatedly calculates the similarity between the pairs,
A data matching method between business models, in which a data processing unit has a process of filtering calculation results and presenting pairs with high similarity on a display device. 2. The matching method according to claim 1, wherein, in the process of normalizing the process model, the data processing unit rewrites the notation into a loop notation when a return side exists in the process model. In the process of normalizing the process model, the data processing unit rewrites the notation into a parallel execution process when there is a notation of a process having no data relation between continuously executed processes. The matching method according to claim 1 or 2. In the process of generating the extended process graph, the data processing unit attaches different labels to edges derived from the data model and edges derived from the process model, respectively. Matching method as described in. Processing for generating the extended process graph when an element having a matching character string is detected between the process model corresponding to the first business model and the process model corresponding to the second business model; 5. The matching according to claim 1, wherein the data processing unit to execute attaches different labels to the upstream side and the downstream side to the element for which a match is detected. Method. 6. The matching according to claim 1, wherein, in the process of generating the pair-wise connection graph, the data processing unit creates a pair in consideration of the order of the normalized process model. Method. In the process of generating the similarity propagation graph, the data processing unit obtains a similarity between texts using a thesaurus or an editing distance for a clause that is a text pair, and uses this as an initial value of the similarity between the pairs. It sets. The matching method of any one of Claims 1-6 characterized by the above-mentioned. In the process of generating the similarity propagation graph, the data processing unit changes an initial value of the propagation coefficient according to whether the edge propagation coefficient is derived from a process model or a data model. Item 8. The matching method according to any one of Items 1 to 7. A process of reading data from the data model and process model corresponding to the first business model, and a data model and process model data corresponding to the second business model, respectively from the storage device;
A character string of each element constituting the process model corresponding to the first business model and a character string of each element constituting the process model corresponding to the second business model are compared on the data, and the matching relationship is A process to detect the allowed elements;
Normalizing the model notation of the process model corresponding to the first and second business models into a notation using a preset model notation;
Based on the relationship between the data model corresponding to the first business model and the elements of the process model, processing for generating on the data a first extended process graph that integrates them into one graph;
Based on the relationship between the data model corresponding to the second business model and the elements of the process model, a process of generating on the data a second extended process graph that integrates these into one graph;
Forming a pair-wise connection graph on the data that forms a pair with the elements of the first extended process graph and the elements of the second extended process graph and that describes the relationship between each pair with a label;
Processing to generate a similarity propagation graph in which the similarity and propagation coefficient between pairs are set on the data;
A process of calculating similarity between pairs by iterative operations;
A computer program that causes a computer to execute a process of filtering a calculation result and presenting a highly similar pair on a display device.

Images