JP2012064136A - Test data generation method, test data generation device, and test data generation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate test data simulating a one-sided link structure in a latent community.SOLUTION: A cluster generation part 11 generates a plurality of clusters by dividing a node group consisting of a plurality of nodes given identification information at random into the number of clusters determined by the total number of nodes and the number of cluster scales. A graph data generation part 12 generates respective graph data of the plurality of clusters by setting link attribute information representing link states between nodes constituting the respective clusters using random numbers and thresholds generated for between the nodes. A test data generation part 13 generates test data simulating the link states between the respective nodes constituting the node group by integrating all the graph data generated by the graph data generation part 12 based upon the identification information by performing control such that the cluster generation processing and graph data generation processing are repeated as many times as multiplex association is possible.

Description

  The present invention relates to a test data generation method, a test data generation device, and a test data generation program.

  Conventionally, the validity of an algorithm used for information processing in which an external input exists is determined. Here, when the external input can be easily acquired as actual data, the actual data can be used as data for testing the validity of the algorithm. For example, as actual data that can be collected, graph data for analyzing a link structure between users in an SNS (Social Networking Service) is generated.

  On the other hand, when actual data cannot be easily collected, test data simulating an external input is required to determine the validity of the algorithm. Collecting actual data for testing in an algorithm whose analysis target is a communication network is that the analysis target is large-scale, and the state of the actual data being collected changes in real time when the actual data is collected sequentially. It may be difficult to do so. That is, it is necessary to generate test data in order to determine the validity of an algorithm whose analysis target is a communication network.

  For this reason, since the advent of the telephone network, a device for generating test data that simulates an external input has been used in testing of an exchange or the like in order to verify a control algorithm of the telephone network. If the communication path is associated with a physical line as in the case of a telephone and the connection structure of the line is obvious from geographical conditions, the test data can be generated relatively easily. That is, since the graph structure can be easily acquired, the test data can be easily generated by determining the communication density with, for example, a random number.

JP 2008-052495 A

  By the way, the concept of connection in the current Internet service also exists in a link between Web pages that is not between physical devices. In addition, direct transition is possible between all Web pages as long as a link is described. Therefore, in the Internet service, it is difficult to extract a trivial connection structure based on geographical conditions as in a telephone network.

  Also, not all Web pages are linked at random with the same probability, but links that are biased due to potential ties, etc. are provided. Furthermore, in search engines, etc., there is a demand for an algorithm that uses the relationship due to a biased link in a potential community. In operation verification such as a search algorithm, if the test data does not simulate a biased link structure in a potential community, the validity of the verification becomes uncertain.

  However, since the conventional technology described above is based on the assumption that the connection structure of the line can be acquired, test data that simulates a biased link structure in a potential community, such as a link between Web pages, is generated. I couldn't.

  Accordingly, the present invention has been made to solve the above-described problems of the prior art, and test data generation that can generate test data that simulates a biased link structure in a potential community. It is an object to provide a method, a test data generation device, and a test data generation program.

  In order to solve the above-described problems and achieve the object, the test data generation method disclosed in the present application is a computer in which a node group composed of a plurality of nodes each having identification information is assigned to the total node group. By randomly dividing the number of nodes and the number of clusters determined by a predetermined number of clusters, a cluster generation step for generating a plurality of clusters and between nodes constituting each cluster generated by the cluster generation step A graph data generation step of generating graph data indicating a link structure in each of the plurality of clusters by setting link attribute information indicating a link state using a random number generated between the nodes and a predetermined threshold value And a cluster generation process by the cluster generation step and the graph data generation step. By controlling the graph data generation process to repeat a predetermined number of times, all the graph data generated by the graph data generation step is integrated based on the identification information given to each node, so that the node group And a test data generation step for generating test data simulating the link state between the nodes constituting the network.

  In addition, the test data generation device disclosed in the present application is configured such that a node group composed of a plurality of nodes each having identification information is assigned to the number of clusters determined by the total number of nodes of the node group and a predetermined cluster size number. By randomly dividing the cluster generation means for generating a plurality of clusters, and link attribute information indicating the link state between the nodes constituting each cluster generated by the cluster generation means, between the nodes A graph data generation unit that generates graph data indicating a link structure in each of the plurality of clusters by setting using the generated random number and a predetermined threshold, and a cluster generation process and the graph data generation by the cluster generation unit By controlling the graph data generation processing by means to be repeated a predetermined number of times, the graph All the graph data generated by the data generation means is integrated based on the identification information given to each node, thereby generating test data that simulates the link state between the nodes constituting the node group. And test data generation means.

  A test data generation program disclosed in the present application causes a computer to execute the test data generation method.

  According to the test data generation method, test data generation apparatus, and test data generation program disclosed in the present application, it is possible to generate test data that simulates a biased link structure in a potential community.

FIG. 1 is a diagram for explaining the configuration of the test data generation apparatus according to the first embodiment. FIG. 2 is a diagram for explaining an overview of processing executed by the test data generation apparatus according to the first embodiment. FIG. 3 is a diagram for explaining the configuration of the test data generation apparatus according to the second embodiment. FIG. 4 is a diagram for explaining the setting list shown in FIG. FIG. 5 is a diagram for explaining the node list shown in FIG. FIG. 6 is a diagram for explaining the data format of the test data shown in FIG. FIG. 7 is a flowchart for explaining a first cluster generation method executed by the cluster generation unit according to the second embodiment. FIG. 8 is a flowchart for explaining a second cluster generation method executed by the cluster generation unit according to the second embodiment. FIG. 9 is a flowchart for explaining processing of the graph data generation unit according to the second embodiment. FIG. 10 is a flowchart for explaining the process of the test data generation apparatus according to the second embodiment. FIG. 11 is a diagram for explaining conventional test data. FIG. 12 is a diagram for explaining test data generated in the second embodiment. FIG. 13 is a diagram illustrating the computer that executes the test data generation program according to the first embodiment.

  Hereinafter, embodiments of a test data generation method, a test data generation device, and a test data generation program disclosed in the present application will be described in detail. Hereinafter, a test data generation apparatus that executes the test data generation method disclosed in the present application will be described as an example. In addition, this invention is not limited by the following examples.

  A test data generation apparatus according to the first embodiment will be described with reference to FIGS. 1 and 2. FIG. 1 is a diagram for explaining the configuration of the test data generation device according to the first embodiment, and FIG. 2 is a diagram for explaining an overview of processing executed by the test data generation device according to the first embodiment. is there.

  As illustrated in FIG. 1, the test data generation device 10 according to the first embodiment includes a cluster generation unit 11, a graph data generation unit 12, and a test data generation unit 13.

  Here, the test data generation device 10 according to the first embodiment, when determining the validity of an algorithm used for information processing in which an external input exists, when the external input cannot be easily acquired as actual data, It is a device that generates simulated test data. For example, the test data generation apparatus according to the first embodiment generates test data for performing operation verification of a Web page search algorithm in an Internet service. Specifically, the test data generation apparatus according to the first embodiment generates test data that simulates a biased link structure in a potential community, such as a link between Web pages.

  More specifically, the cluster generation unit 11, the graph data generation unit 12, and the test data generation unit 13 cooperate with each other when an operator of the test data generation apparatus 10 receives a setting list as shown in FIG. To generate test data. Here, as shown in FIG. 2A, the total number of nodes (n), the number of cluster scales (m), the number of possible multiple assignments (t), and the threshold (s) are set in the setting list. Hereinafter, processing of the cluster generation unit 11, the graph data generation unit 12, and the test data generation unit 13 executed using each numerical value described in the setting list will be described.

  The cluster generation unit 11 first sets a node group composed of a plurality of nodes to which identification information is assigned.

  For example, as illustrated in FIG. 2A, when “total number of nodes (n): 4” is set, the cluster generation unit 11 sets a node group including four nodes. Here, the cluster generation unit 11 assigns node numbers “1 to 4” as identification information to each of the four nodes. As a result, as shown in FIG. 2A, the cluster generation unit 11 has a node group “N = {1, 2, 3, 3] composed of four nodes assigned node numbers“ 1 to 4 ”. 4} ". The identification information may be any information as long as it can uniquely identify each node, and may be a name, for example.

  Here, when the total number of nodes “n” is set, the test data generation unit 13 generates test data simulating the link state between “n × n” nodes. For example, the test data generation unit 13 generates test data by setting a link state between nodes in each element of an “n × n” square matrix. Here, the test data generation unit 13 initializes the test data to an “n × n” zero matrix or unit matrix.

  That is, when the operator wants to generate test data that simulates the strength of the relationship between nodes, it is appropriate to set that each node is completely connected to its own node. The test data generation unit 13 initializes the test data to be generated in an “n × n” unit matrix. In addition, when the operator wants to generate test data that simulates the presence or absence of a link between nodes, it is appropriate to set the own node not to be a link destination. The test data to be generated is initialized to a zero matrix of “n × n”.

  For example, in response to an instruction from an operator who wants to generate test data simulating the strength of relevance between nodes, the test data generation unit 13 generates test data “Ea” to be generated as shown in FIG. Initially set to a “4 × 4” unit matrix.

  Then, the cluster generation unit 11 generates a plurality of clusters by randomly dividing the node group into the number of clusters determined by the total number of nodes (n) and the number of cluster sizes (m). For example, when “m” is a divisor of “n”, the cluster generation unit 11 performs “c = n / m” times from “n” nodes constituting the node group to “m” nodes. Are randomly selected. Accordingly, the cluster generation unit 11 divides the node group including “n” nodes into clusters “M (1) to M (c)” including “m” nodes.

  In the example shown in FIG. 2A, the cluster generation unit 11 is set as “total number of nodes (n): 4, cluster size number (m): 2” in the setting list, so “2 = The nodes of “2” nodes are sequentially selected at random from “4” nodes constituting the node group over 4/2 times. Thereby, the cluster generation unit 11 divides the node group “N” composed of “4” nodes into clusters “M (1) and M (2)” composed of “2” nodes.

  In the example illustrated in FIG. 2B, the cluster generation unit 11 sets the node group “N” to “M (1) = {1, 3}” and “M (2) = {2, 4}”. To divide. The processing of the cluster generation unit 11 when the total number of nodes is not divisible by the number of cluster scales will be described in detail later.

  Then, the graph data generation unit 12 shown in FIG. 1 uses the link attribute information indicating the link state between the nodes constituting each cluster generated by the cluster generation unit 11 as a random number and a threshold generated between the nodes. By setting using (s), graph data indicating a link structure in each of a plurality of clusters is generated.

  Specifically, when setting the link attribute information “e (i, j)” from the node number “i” to the node number “j”, the graph data generation unit 12 first generates a random number “r”. To do. Then, for example, when the random number “r” is larger than the threshold value “s”, the graph data generating unit 12 sets the link attribute information based on “r” as “e (i, j)”, and the random number “ If “r” is equal to or less than the threshold “s”, “e (i, j)” is set to “0”. The graph data generation unit 12 sets the link attribute information “e (i, i)” from the node number “i” to the node number “i” as “0”.

  For example, in the cluster “M (1)”, the graph data generation unit 12 sets link attribute information from the node number “1” to the node number “3” as shown in FIG. A random number “r” is generated, and “e (1,3)” is set based on the magnitude relationship between “r” and “s”. Further, as shown in FIG. 2B, the graph data generating unit 12 sets random numbers “r” and “r” for setting link attribute information from the node number “3” to the node number “1”. Based on the magnitude relationship with “s”, “e (3,1)” is set. As a result, the graph data generation unit 12 generates a square of “2 × 2” in which the link attribute information is set for each element as the graph data of the cluster “M (1)” as illustrated in FIG. A matrix “E1” is generated.

  Further, as shown in FIG. 2B, the graph data generation unit 12 sets link attribute information from the node number “2” to the node number “4” in the cluster “M (2)”. A random number “r” is generated and “e (2, 4)” is set, and a random number “r” for setting link attribute information from the node number “4” to the node number “2” is generated and “ e (4,2) "is set. As a result, the graph data generation unit 12 generates a square of “2 × 2” in which the link attribute information is set for each element as graph data of the cluster “M (2)”, as shown in FIG. A matrix “E2” is generated. The order in which the link attribute information is set may be any order.

  Then, the test data generation unit 13 illustrated in FIG. 1 performs control so that the cluster generation processing by the cluster generation unit 11 and the graph data generation processing by the graph data generation unit 12 are repeated so as to repeat the number (t) of multiple assignments. All the graph data generated by the generation unit 12 is integrated based on the identification information given to each node, thereby generating test data that simulates the link state between the nodes constituting the node group.

  For example, as shown in FIG. 2B, the test data generation unit 13 incorporates E1 and E2 generated by the graph data generation unit 12 in the first process into the initial setting Ea according to the node number. Then, the test data generation unit 13 sends an instruction to the cluster generation unit 11 and the graph data generation unit 12 to perform the second process.

  Thereby, as shown in FIG. 2C, the cluster generation unit 11 sets the node group “N” to “M (1) = {1, 4}” and “M” in the second process. (2) = {2, 3} ”is divided into two clusters. Then, the graph data generation unit 12 sets link attribute information from the node number “1” to the node number “4” in the cluster “M (1)”, as shown in FIG. A random number “r” is generated and “e (1, 4)” is set, and a random number “r” for setting link attribute information from the node number “4” to the node number “1” is generated and “ e (4,1) "is set. As a result, the graph data generation unit 12 generates “E1” as the second graph data of the cluster “M (1)”, as shown in FIG.

  Further, as shown in FIG. 2C, the graph data generation unit 12 sets link attribute information from the node number “2” to the node number “3” in the cluster “M (2)”. A random number “r” is generated and “e (2, 3)” is set, and a random number “r” for setting link attribute information from the node number “3” to the node number “2” is generated and “ e (3,2) "is set. As a result, the graph data generation unit 12 generates “E2” as the second graph data of the cluster “M (2)”, as shown in FIG.

  Then, the test data generation unit 13 incorporates E1 and E2 generated by the second process according to the node number with respect to Ea in which E1 and E2 generated by the first process are installed. As a result, the test data generation unit 13 generates Ea reflecting the first and second processing results of the graph data generation unit 12, as shown in FIG.

  Then, the test data generation unit 13 repeats the processes of the cluster generation unit 11 and the graph data generation unit 12 “t” times as illustrated in FIG. Accordingly, the test data generation unit 13 generates test data “Ea” in which all the graph data generated by the graph data generation unit 12 through the first to t-th processing is integrated. Note that, as described above, the graph data integration process by the test data generation unit 13 may be executed every time the process of the graph data generation unit 12 is completed, or the graph data generation unit 12 t It may be a case where batch processing is executed after batch processing is completed.

  As described above, according to the first embodiment, the cluster generation unit 11 determines a node group including a plurality of nodes to which identification information is assigned, based on the total number of nodes and the number of cluster sizes of the node group. Multiple clusters are generated by randomly dividing the number of clusters. Then, the graph data generation unit 12 sets the link attribute information indicating the link state between the nodes constituting each cluster generated by the cluster generation unit 11 using the random numbers and threshold values generated between the nodes. Thus, graph data indicating the link structure in each of the plurality of clusters is generated. Then, the test data generation unit 13 generates the graph data generation unit 12 by controlling the cluster generation processing by the cluster generation unit 11 and the graph data generation processing by the graph data generation unit 12 to repeat the number of multiple assignments. By integrating all the graph data based on the identification information given to each node, test data simulating the link state between each node constituting the node group is generated.

  That is, in the first embodiment, test data is generated by repeating the cluster generation process and the graph data generation process a plurality of times, and further integrating all the generated graph data. Here, the cluster generated by the cluster generation unit 11 can be regarded as a potential community. The graph data generated by the graph data generation unit 12 can be regarded as data simulating a biased link structure in the community. That is, the data generated in the first embodiment is data that simulates a communication network in which a plurality of communities having a biased link structure are formed in an intricate manner. Therefore, in the first embodiment, it is possible to generate test data that simulates a biased link structure in a potential community.

  In the second embodiment, the cluster generation process, the graph data generation process, and the test data generation process described in the first embodiment will be described in detail.

  First, the test data generation apparatus 10 according to the second embodiment will be described with reference to FIG. FIG. 3 is a diagram for explaining the configuration of the test data generation apparatus according to the second embodiment.

  As illustrated in FIG. 3, the test data generation apparatus according to the second embodiment includes an input unit 20, an output unit 30, an input / output control I / F unit 40, a storage unit 50, and a processing unit 60.

  The input unit 20 includes a reading device for a recording medium such as a mouse, a keyboard, an FD (Flexible Disk Drive), a CD-ROM (Compact Disc Read Only Memory), an MO (Magneto Optical Disc), and a DVD (Digital Versatile Disc). Then, various setting information is received from the operator of the test data generation apparatus 10. Specifically, the input unit 20 is a setting in which the total number of nodes (n), the number of cluster sizes (m), the number of possible multiple assignments (t), and the threshold value (s) are set by the operator of the test data generation apparatus 10. Accept registration of list.

  The output unit 30 includes a monitor, a speaker, and the like, and displays, for example, a processing result of the processing unit 60 described later on the monitor.

  The input / output control I / F unit 40 controls data transfer between the input unit 20 and the output unit 30, and the storage unit 50 and the processing unit 60.

  The storage unit 50 stores data used for various processes by the processing unit 60 and various processing results by the processing unit 60. The storage unit 50 is, for example, an HDD (Hard Disc Drive) or a RAM (Random Access Memory). Specifically, as shown in FIG. 3, the storage unit 50 includes a setting list 51, a node list 52, cluster-specific graph data 53, and test data 54.

  The setting list 51 stores the setting list received by the input unit 20. FIG. 4 is a diagram for explaining the setting list shown in FIG.

  For example, as shown in FIG. 4A, the setting list 51 stores a setting list in which “total number of nodes: n, number of multiple assignments: t, number of cluster sizes: m, threshold: s” is set. To do. Alternatively, as shown in FIG. 4B, for example, the setting list 51 includes “m1, m2” in which “total number of nodes: n, number of possible multiple assignments: t” and t cluster scale numbers are arranged. ,..., Mt ”and“ s1, s2,..., St ”in which t threshold values are arranged are stored.

  That is, in the example shown in FIG. 4A, all cluster generation processes repeated t times are executed with the same number of cluster scales, and all graph data generation processes repeated t times are executed with the same threshold value. Is set to On the other hand, in the example shown in FIG. 4B, the number of cluster sizes used in the cluster generation process repeated t times is set to change, and the threshold value used in the graph data generation process repeated t times changes. Is set to Each of the t cluster scale numbers may be different from each other, or a part or all of them may have the same value. Similarly, each of the t threshold values may be different from each other, or may be a case where a part or all of them are the same value.

  The node list 52 stores a list of a plurality of nodes constituting the node group set by the cluster generation unit 11 of the processing unit 60 based on the total number “n” of the setting lists stored in the setting list 51. The area size of the node list 52 in the storage unit 50 is ensured by, for example, the cluster generation unit 11 according to the total number of nodes. FIG. 5 is a diagram for explaining the node list shown in FIG.

  In the node list 52, for example, as shown in FIG. 5A, node numbers “1, 2, 3,..., N−1, n” are assigned to each of “n” nodes. Is stored as a node list in the format assigned as. Alternatively, the node list 52 includes, for example, the names “name (1), name (2), name (3),..., Name (n−1), name as shown in FIG. (N) "is stored as a node list of identification information. A case will be described below where the node list 52 stores a node list in a format in which a node number is assigned as identification information to each of “n” nodes.

  Returning to FIG. 3, the cluster-specific graph data 53 includes a list of nodes constituting each cluster generated by the cluster generation unit 11 of the processing unit 60 and a link list of each cluster generated by the graph data generation unit 12 of the processing unit 60. Remember. That is, the cluster-specific graph data 53 stores a cluster-specific node list in which a list of nodes constituting each cluster generated by the cluster generation unit 11 is grouped by cluster. Further, the cluster-specific graph data 53 stores a cluster-specific link list in which a list of link attribute information between nodes constituting each cluster generated by the cluster generation unit 11 is compiled for each cluster. In other words, the cluster-specific graph data 53 stores the graph data of each cluster by storing the cluster-specific node list and the cluster-specific link list as a pair for each cluster.

  The area size in the storage unit 50 of the cluster-specific node list is determined by, for example, the cluster generation unit 11 and the graph data generation unit 12 according to the total number of nodes, the number of cluster sizes, the number of possible multiple assignments, and the cluster generation method described later Secured. That is, the area scale of the cluster-specific node list is the area scale obtained by adding the number of nodes constituting each cluster, which is repeatedly generated by the cluster generation unit 11 as many times as possible. The area scale of the cluster-specific link list is an area scale obtained by adding the squares of the number of nodes constituting each cluster generated by the cluster generation unit 11 repeatedly for the number of multiple assignments.

  The test data 54 stores test data generated by the test data generation unit 13 of the processing unit 60. When the total number of nodes “n” is set, the test data simulating the link state between the nodes becomes data simulating the link state between “n × n” nodes. That is, the area size of the test data 54 in the storage unit 50 is secured by, for example, the test data generation unit 13 as an area size corresponding to the square of the total number of nodes. FIG. 6 is a diagram for explaining the data format of the test data shown in FIG.

  For example, as shown in FIG. 6A, the test data 54 stores test data in a data format of an “n × n” square matrix. In the example shown in FIG. 6A, the element in the i-th row and j-th column is the link attribute information from the i-th node as the link source to the j-th node as the link destination.

  Alternatively, the test data 54 stores test data in a data format of n pieces of array information such as “link source, link destination, link attribute information” as shown in FIG. In the following, a case will be described in which the test data 54 is stored in the “n × n” square matrix data format shown in FIG.

  Returning to FIG. 3, the processing unit 60 includes a cluster generation unit 11, a graph data generation unit 12, and a test data generation unit 13 in order to generate test data based on the setting list stored in the setting list 51.

  First, the cluster generation unit 11 sets a node group composed of a plurality of nodes to which identification information is assigned, and stores the set node group in the node list 52 as a node list. For example, the cluster generation unit 11 stores the node list illustrated in FIG.

  Then, the cluster generation unit 11 generates a plurality of clusters by randomly dividing the set node group into the number of clusters determined by the total node number (n) and the cluster size number (m). The cluster generation unit 11 stores the generated list of nodes constituting each cluster in the cluster-specific graph data 53 as a cluster-specific node list.

  Here, when “m” is a divisor of “n”, the cluster generation unit 11 performs “c = n / m” times from “n” nodes constituting the node group to “m” pieces. Select nodes sequentially and randomly. Accordingly, the cluster generation unit 11 divides the node group including “n” nodes into clusters “M (1) to M (c)” including “m” nodes.

  However, if “m” is not a divisor of “n”, the cluster generation unit 11 needs to process as many nodes as the remainder obtained by dividing “n” by “m”. Therefore, the cluster generation unit 11 according to the second embodiment performs cluster generation processing using a first cluster generation method or a second cluster generation method described below.

  First, in both the first cluster generation method and the second cluster generation method, the cluster generation unit 11 according to the second embodiment generates a cluster size number randomly from a node group that is not extracted as a cluster among a plurality of nodes. Clusters are generated sequentially by extracting nodes.

  Here, when “m” is a divisor of “n”, the cluster generation unit 11 according to the second embodiment executes the cluster generation processing only by the above processing. On the other hand, if “m” is not a divisor of “n” and the above process is repeated “integer number of times that is not more than a value obtained by dividing the total number of nodes in the node group by the number of cluster sizes”, the cluster is not extracted The number of nodes in the node group is smaller than the cluster scale “m”.

  Therefore, the cluster generation unit 11 according to the second embodiment, as the first cluster generation method, when the number of nodes in the unextracted node group as a cluster is smaller than the number of cluster sizes, the unextracted node group as the cluster is selected. Randomly allocate to each generated cluster. As a result, the cluster generation unit 11 according to the second embodiment executes the first cluster generation method to obtain an integer number of clusters that is the maximum of the number of nodes less than the value obtained by dividing the total number of nodes in the node group by the number of cluster sizes. Generate.

  Alternatively, the cluster generation unit 11 according to the second embodiment, as the second cluster generation method, when the number of nodes in the node group that has not been extracted as a cluster is smaller than the number of cluster sizes, the node group that has not been extracted as the cluster. Create as a new cluster. As a result, the cluster generation unit 11 according to the second embodiment executes the second cluster generation method, so that an integer number of clusters that are the minimum number of clusters equal to or greater than the value obtained by dividing the total number of nodes in the node group by the number of cluster sizes is obtained. Generate.

  Hereinafter, each of the first cluster generation method and the second cluster generation method will be described in detail with reference to the flowcharts of FIGS. 7 and 8. FIG. 7 is a flowchart for explaining a first cluster generation method executed by the cluster generation unit according to the second embodiment. FIG. 8 is a second cluster generation executed by the cluster generation unit according to the second embodiment. It is a flowchart for demonstrating a method.

  When executing the first cluster generation method, as shown in FIG. 7, the cluster generation unit 11 refers to the setting list, sets the node group “N” of the total number of nodes “n”, and sets the number of cluster sizes. “M” is acquired (step SA1). Here, the cluster generation unit 11 prepares node groups N (0) to N (C) as internal processing data. When the first cluster generation method is executed, “C” is a value obtained by subtracting “1” from an integer that is equal to or less than a value obtained by dividing the total node number “n” by the cluster size number “m”.

  Then, the cluster generation unit 11 sets “N (0) = N”, sets the cluster management variable “i” to “0” (step SA2), and randomly selects m elements from N (i). The element is M (i + 1) (step SA3).

  Thereafter, the cluster generation unit 11 increments “i” to “i = i + 1” (step SA4) and “N (i) = N (i−1) −M (i)” (step SA5). ).

  Then, the cluster generation unit 11 determines whether the number of elements of N (i) is greater than or equal to m (step SA6). Here, when the number of elements of N (i) is greater than or equal to m (Yes at step SA6), the cluster generation unit 11 returns to step SA3 and selects elements that form the cluster M (i + 1) from N (i). Perform processing.

  On the other hand, when the number of elements of N (i) is smaller than m (No at Step SA6), the cluster generation unit 11 sets the residual node distribution management variable (J) to “J = 0” (Step SA7). One element is selected at random from N (i) and added to M (J + 1), and the added element is removed from N (i) (step SA8).

  Then, the cluster generation unit 11 sets “J = (J + 1) mod (maximum integer not exceeding n / m)” (step SA9), and determines whether the number of elements of N (i) is “0”. (Step SA10). Here, when the number of elements of N (i) is not “0” (No at Step SA10), the cluster generation unit 11 returns to Step SA8 and performs distribution processing of the surplus nodes.

  On the other hand, when the number of elements of N (i) becomes “0” (Yes at step SA10), the cluster generation unit 11 ends the process.

  When executing the second cluster generation method, as illustrated in FIG. 8, the cluster generation unit 11 refers to the setting list, sets the node group “N” of the total number of nodes “n”, and The scale number “m” is acquired (step SB1). Here, the cluster generation unit 11 prepares node groups N (0) to N (C) as internal processing data. When the second cluster generation method is executed, “C” is an integer that becomes the maximum value not more than a value obtained by dividing the total node number “n” by the cluster size number “m”.

  Then, the cluster generation unit 11 sets “N (0) = N”, sets the cluster management variable “i” to “0” (step SB2), and selects m elements at random from N (i). The element is M (i + 1) (step SB3).

  Thereafter, the cluster generation unit 11 increments “i” to “i = i + 1” (step SA4) and “N (i) = N (i−1) −M (i)” (step SB5). ).

  Then, the cluster generation unit 11 determines whether the number of elements of N (i) is greater than or equal to m (step SB6). Here, when the number of elements of N (i) is greater than or equal to m (Yes in step SB6), the cluster generation unit 11 returns to step SB3 and selects elements that constitute the cluster M (i + 1) from N (i). Perform processing.

  On the other hand, when the number of elements of N (i) is smaller than m (No at Step SB6), the cluster generation unit 11 sets “M (i + 1) = N (i)” (Step SB7), and ends the process. .

  That is, the first cluster generation method is a method of distributing the remaining number of nodes obtained by dividing “n” by “m” to the generated clusters (number of elements: m). On the other hand, in the second cluster generation method, the remaining number of nodes obtained by dividing “n” by “m” are used as new clusters as they are. Whether to use the first cluster generation method or the second cluster generation method gives priority to equalizing the number of nodes constituting each cluster as much as possible, or “m” as many as possible. Depends on whether to give priority to not changing. The operator selects either the first cluster generation method or the second cluster generation method according to the purpose of the test target algorithm using the test data. For example, whether the cluster generation unit 11 executes the first cluster generation method or the second cluster generation method is determined by the operator when registering the setting list or before starting the processing of the cluster generation unit 11. It is specified.

  Returning to FIG. 3, the graph data generation unit 12 uses the link attribute information indicating the link state between the nodes constituting each cluster generated by the cluster generation unit 11, the random number and threshold value generated between the nodes. By setting using (s), graph data indicating a link structure in each of a plurality of clusters is generated. Then, the graph data generation unit 12 associates the list of link attribute information between the nodes set in the graph data generated from each cluster as a cluster-specific link list, and associates the list with the cluster-specific node list of the cluster to be generated. This is stored in the graph data 53 for each cluster.

  Here, the graph data generation part 12 sets the weight which shows the presence or absence of the link between nodes, or the relationship strength between nodes as link attribute information which shows the link state between nodes. The type of link attribute information is determined by an instruction from the operator. Below, the case where the graph data generation part 12 sets the weight which shows the relationship strength between nodes as link attribute information is demonstrated.

  Specifically, when setting the weighting “e (i, j)” from the node number “i” to the node number “j”, the graph data generation unit 12 first generates a random number “r”. For example, the graph data generation unit 12 generates a real number in the range of “0 to 1” as a random number. The graph data generation unit 12 sets “e (i, j)” to “0” when the random number “r” is equal to or less than the threshold value “s”. Further, the graph data generation unit 12 sets the weight “e (i, j)” to “0” when “i = j”.

  On the other hand, when “i = j” is not satisfied and the random number “r” is larger than the threshold “s”, the graph data generation unit 12 sets “e (i, j)” based on “r”. . Specifically, the graph data generation unit 12 sets “e (i, j)” by any one of the three setting methods described below.

  The first setting method is a method of setting the random number “r” as it is as “e (i, j)” if “r> s”. The second setting method is a method of setting “r−s” as “e (i, j)” if “r> s”. The third setting method is a method of setting “(rs− / 1-s)” as “e (i, j)” if “r> s”. Note that the graph data generation unit 12 sets weights between nodes by a setting method designated by the operator from the first to third setting methods. For example, whether the graph data generating unit 12 executes the first setting method, the second setting method, or the third setting method depends on whether the graph data generating unit 12 performs processing when registering the setting list. It is specified by the operator before starting.

  When the graph data generation unit 12 sets the presence / absence of a link between nodes as the link attribute information, for example, if the random number “r” is larger than the threshold “s”, “e (i, j)” Is set to “with link”, and if the random number “r” is equal to or less than the threshold value “s”, “e (i, j)” is set to “without link”. The graph data generation unit 12 sets “e (i, j)” to “no link” when “i = j”.

  Hereinafter, the process of the graph data generation part 12 is demonstrated using FIG. FIG. 9 is a flowchart for explaining processing of the graph data generation unit according to the second embodiment. FIG. 9 illustrates a case where the graph data generation unit 12 sets the weight indicating the relationship strength between nodes as the link attribute information by the second setting method.

  As illustrated in FIG. 9, the graph data generation unit 12 according to the second embodiment refers to the cluster-specific node list and setting list 51 of the cluster-specific graph data 53, and includes the cluster having the number of elements “y” and the threshold value “s”. Is acquired (step SC1).

  Then, the graph data generation unit 12 sets the link source node number management variable (i) and the link destination node number management variable (j), which are the link arrival and departure node numbers, to “i = 1, j = 1”, and the cluster A matrix “E = [y × y]” for storing another graph data is initialized to a zero matrix (step SC2).

  Thereafter, the graph data generation unit 12 determines whether or not “i = j” (step SC3). Here, if “i = j” (Yes at step SC3), the graph data generation unit 12 sets the weight “e (i, j)” from the node number “i” to the node number “j” to “0”. Is set (step SC7).

  When it is not “i = j” (No at Step SC3), the graph data generation unit 12 generates a random number (r) within the range of “0 to 1” (Step SC4). Then, the graph data generating unit 12 determines whether or not “r> s” (step SC5). Here, when “r” is equal to or less than “s” (No in step SC5), the graph data generating unit 12 sets “e (i, j)” to “0” (step SC7).

  On the other hand, if “r> s” (Yes at step SC5), the graph data generating unit 12 sets “e (i, j)” to “rs” (step SC6). When the first setting method is performed, the graph data generation unit 12 sets “e (i, j)” to “r”. When the third setting method is performed, the graph data generation unit 12 sets “e (i, j)” to “(rs) / (1-s)”.

  After performing the process of step SC6 or the process of step SC7, the graph data generating unit 12 increments “j” to “j = j + 1” (step SC8), and whether “j> y” is satisfied. Is determined (step SC9). Here, when “j” is equal to or less than “y” (No at Step SC9), the graph data generating unit 12 returns to Step SC3 and determines whether or not “i = j”.

  On the other hand, if “j> y” (Yes at step SC9), the graph data generation unit 12 increments “i” to “i = i + 1”, resets “j” to “j = 1”. (Step SC10), it is determined whether or not “i> y” (step SC11). Here, when “i” is equal to or less than “y” (No in step SC11), the graph data generating unit 12 returns to step SC3 and performs a determination process as to whether or not “i = j”.

  On the other hand, if “i> y” (Yes at step SC11), the graph data generating unit 12 ends the process.

  In FIG. 9, the case where the link source node number management variable (i) is incremented after the link destination node number management variable (j) is incremented has been described. However, the graph data generation process may be a case where the link destination node number management variable (j) is incremented after the link source node number management variable (i) is incremented.

  Returning to FIG. 3, the test data generation unit 13 performs control so that the cluster generation process by the cluster generation unit 11 and the graph data generation process by the graph data generation unit 12 are repeated (t). Then, the test data generation unit 13 integrates all the graph data generated by the graph data generation unit 12 based on the identification information given to each node, so that the links between the nodes constituting the node group Generate test data that simulates the condition. Then, the test data generation unit 13 stores the generated test data in the test data 54.

  Specifically, when the link attribute information is a weight indicating the relationship strength between the nodes, the test data generation unit 13 obtains the test data by calculating the arithmetic sum of all the graph data generated by the graph data generation unit 12. Generate. For example, the test data generation unit 13 generates all the graph data (relationship strength between nodes) generated by the graph data generation unit 12 with respect to the test data (Ea) that is initially set in the “n × n” unit matrix. Test data is generated by calculating the arithmetic sum.

  Note that the test data generation unit 13 generates test data by taking the logical sum of all the graph data generated by the graph data generation unit 12 when the link attribute information is the presence or absence of a link between nodes. For example, the test data generation unit 13 performs a logical sum of all the graph data (the presence / absence of a link) generated by the graph data generation unit 12 with respect to the test data (Ea) initially set to an “n × n” zero matrix. Test data is generated by taking

  Here, the test data generation unit 13 may change at least one of the cluster size number and the threshold every time the cluster generation process and the graph data generation process are executed (t). For example, as shown in FIG. 4B, when the cluster size number used for the cluster generation process repeated t times is set to change, the cluster generation unit 11 controls the test data generation unit 13. Thus, the cluster size number corresponding to the number of repetitions actually executed is acquired from the setting list, and the cluster generation processing illustrated in FIG. 7 or FIG. 8 is performed.

  In addition, as illustrated in FIG. 4B, when the setting list is set so that the threshold used for the graph data generation process repeated t times is stored, the graph data generation unit 12 performs the test Under the control of the data generation unit 13, a threshold corresponding to the number of repetitions to be executed is acquired from the setting list, and the graph data generation process illustrated in FIG. 9 is performed. As shown in FIG. 4B, the cluster size number and the threshold value may be set for each repetition count, or one of them is a single value. There may be.

  In addition, the cluster generation unit 11 executes either the first cluster generation method or the second cluster generation method for each repetition count under the control of the test data generation unit 13 based on an instruction from the operator, for example. It may be the case. For example, for each iteration, whether the cluster generation unit 11 executes the first cluster generation method or the second cluster generation method depends on whether the setting list is registered or before the cluster generation unit 11 starts processing. Specified by the operator. Further, when setting the weighting, the graph data generation unit 12 controls the test data generation unit 13 based on an instruction from the operator, for example, for each repetition count, the first setting method and the second setting method. Alternatively, any one of the third setting methods may be executed. For example, for each repetition count, whether the graph data generation unit 12 executes the first setting method, the second setting method, or the third setting method is determined when the setting list is registered or the graph data It is designated by the operator before the generation unit 12 starts processing.

  Subsequently, a processing procedure performed by the test data generation apparatus 10 according to the second embodiment will be described with reference to FIG. FIG. 10 is a flowchart for explaining the process of the test data generation apparatus according to the second embodiment. FIG. 10 is a flowchart for explaining processing when the link attribute information is weighting indicating the relationship strength between nodes.

  As illustrated in FIG. 10, the test data generation device 10 according to the second embodiment determines whether or not a setting list is stored in the setting list 51 (Step S <b> 101). Here, when the setting list is not stored (No at Step S101), the test data generating apparatus 10 enters a standby state.

  On the other hand, when the setting list is stored (Yes in step S101), the test data generation unit 13 sets the multiple assignable number management variable (k) to “k = 0” and stores “n ×” for storing the test data. The square matrix (Ea) of “n” is set to “Ea = unit matrix” (step S102). If the link attribute information is the presence or absence of a link between nodes, the test data generation unit 13 sets “Ea = zero matrix” in step S102.

  Then, the cluster generation unit 11 performs a cluster generation process (step S103). Specifically, the cluster generation unit 11 performs the first cluster generation process described with reference to FIG. 7 or the second cluster generation process described with reference to FIG.

  Thereafter, the test data generation unit 13 sets the cluster number management counter variable (L) to “L = 0” (step S104), and the graph data generation unit 12 performs a graph data generation process (step S105). For example, when the weighting setting method is set as the first setting method, the graph data generation unit 12 performs the graph data generation processing described with reference to FIG.

  Then, the test data generation unit 13 increments “L” to “L = 1” (step S106), and determines whether “L” is equal to or greater than the number of clusters generated in step S103 (step S106). S107). Here, when “L” is smaller than the number of clusters generated in step S103 (No in step S107), the test data generation unit 13 determines that there is a cluster in which the graph data has not been generated, and the graph data generation unit 12 Under the control of the test data generation unit 13, the process returns to step S105 to perform graph data generation processing for the remaining clusters.

  On the other hand, when “L” is equal to or greater than the number of clusters generated in step S103 (Yes in step S107), the test data generation unit 13 sets “k = k + 1” (step S108), and sets the graph data group of k to “E”. (Step S109). That is, when “L” is equal to or greater than the number of clusters generated in step S103, the test data generation unit 13 determines that the graph data for the number of clusters generated in step S103 has been generated.

  Then, the test data generation unit 13 updates the test data as “Ea = Ea + E” (step S110), and determines whether or not “k = t (multiple assignment possible number)” (step S111).

  Here, when “k” has not reached “t” (No in step S111), the test data generation unit 13 determines that the graph data group corresponding to the number of multiple assignments has not been generated, and the cluster generation unit 11 returns to step S103 under the control of the test data generation unit 13, and performs the cluster generation process again.

  On the other hand, if “k = t” (Yes at Step S111), the test data generation unit 13 sets “Ea” updated at Step S110 as test data, and ends the process.

  In FIG. 10, the case where “Ea” is sequentially updated has been described. However, in the second embodiment, after graph data groups corresponding to the number of possible multiple assignments are generated, they are integrated in a lump. Ea ”may be generated. Further, FIG. 10 illustrates the case where the cluster generation process and the graph data generation process for the number of multiple assignments are sequentially performed. However, in the second embodiment, for example, the cluster generation process and the graph data for the number of multiple assignments can be performed. The generation process may be performed in parallel. FIG. 10 illustrates the case where the graph data generation unit 12 generates graph data for each cluster under the control of the test data generation unit 13 using the cluster number management counter variable (L). 2 may be a case where the graph data generation processing for the plurality of clusters generated in step S103 is performed in parallel.

  As described above, according to the second embodiment, as in the first embodiment, test data is generated by the processing of the cluster generation unit 11, the graph data generation unit 12, and the test data generation unit 13.

  For example, in a telephone network, a communication path is associated with a physical line, and the connection structure of the line is obvious from geographical conditions. Therefore, the test data for verifying the control algorithm of the telephone network simulates the flow of information corresponding to the communication path, for example, by determining the communication density with a random number as shown in FIG. Could be generated as a graph structure. Further, in the test data generation method using a random graph, after giving the possibility that all nodes are connected, the presence or absence of connection is determined by random numbers as shown in FIG. The connection was given a weight. However, the method shown in FIG. 11 cannot generate test data that simulates a biased link structure in a potential community, such as a link between Web pages. FIG. 11 is a diagram for explaining conventional test data.

  However, in the second embodiment, as shown in FIG. 12A, a process for generating a cluster from a node group is performed a plurality of times, and further, as shown in FIG. For each generation process, a graph structure (graph data) of each cluster is generated. And in Example 2, as shown to (C) of FIG. 12, the graph structure of each cluster is integrated. That is, unlike the conventional test data shown in FIG. 11, the test data generated in the second embodiment is data simulating a communication network in a state where a plurality of communities having a biased link structure are formed in an intricate manner. . FIG. 12 is a diagram for explaining test data generated in the second embodiment.

  Therefore, in the second embodiment, as in the first embodiment, it is possible to generate test data that simulates a biased link structure in a potential community.

  In the second embodiment, the graph data generation unit 12 sets the weight indicating the presence / absence of a link between nodes or the relationship strength between nodes as the link attribute information. Then, when the link attribute information is the presence / absence of a link between nodes, the test data generation unit 13 generates test data by taking the logical sum of all the graph data. Further, when the link attribute information is a weight indicating the relationship strength between nodes, the graph data generation unit 12 generates test data by calculating the arithmetic sum of all the graph data.

  Therefore, in the second embodiment, depending on the type desired by the operator, test data simulating the presence or absence of a biased link in a potential community, or a test simulating a biased link density in a potential community. Data can be generated.

  Further, in the second embodiment, the cluster generation unit 11 performs cluster extraction by randomly extracting nodes having a cluster size from a group of nodes not extracted as a cluster among a plurality of nodes as a first cluster generation method. If the number of nodes in a node group that has not been extracted as a cluster is smaller than the number of cluster sizes, the total number of nodes can be determined by randomly allocating the node group that has not been extracted as the cluster to each generated cluster. Generates an integer number of clusters that is the maximum value less than or equal to the number divided by the cluster size number. Alternatively, as a second cluster generation method, the cluster generation unit 11 sequentially generates clusters by randomly extracting nodes of a cluster scale number from a group of nodes not extracted as a cluster among a plurality of nodes. When the number of nodes in a node group that has not been extracted as a cluster is less than the number of cluster sizes, a value obtained by dividing the total number of nodes by the number of cluster sizes by generating a new cluster of nodes that have not been extracted as the cluster. The minimum number of clusters is generated as described above.

  Therefore, in the second embodiment, an operator's request whether to give priority to equalizing the number of nodes constituting each cluster as much as possible or not giving priority to changing the number of nodes constituting each cluster from the number of cluster sizes as much as possible. In response to this, cluster generation processing can be performed. In the second embodiment, the first cluster generation method and the second cluster generation method can be changed every time the cluster generation process is performed.

  In the second embodiment, the test data generation unit 13 performs at least one of the cluster size number and the threshold value every time the cluster generation process and the graph data generation process are executed in accordance with the setting value of the setting list. Change one or the other. That is, in the second embodiment, it is possible to increase the diversity of the assumed community. Therefore, in the second embodiment, it is possible to arbitrarily change the diversity of test data simulating a biased link structure in a potential community.

  That is, when using the test data generation device 10, the operator can generate the test data desired by himself by adjusting the value set in the setting list according to the property of the target to be simulated. it can. For example, when targeting the density (relationship strength) between sites on the Internet, values set by the operator in the setting list are as follows.

  First, the multiple attribution possible number is set as a value representing the assumed number of communities to which a certain site is accessed. When targeting a communication network including many personal sites with a large absolute number on the Internet, the number of multiple belongings can be estimated and set from the number of fields of interest that an individual can have. For example, even if a society with many very hobby people is assumed, the scope of interest at the same time is limited by the structure of the human brain. Recent knowledge about brain structure has revealed that human short-term memory can only hold up to 10 numbers.

  Here, the amount of information that can be stored for a short time is similar to the number of cases that are referred to in a result list in a practically used search system. For this reason, even if there is wide potential interest, there are often about 10 links placed in a prominent place in actual site creation. In view of such properties, it is desirable that the number of possible multiple assignments is set to “10”. Alternatively, depending on the site group to be simulated, the number of possible multiple assignments may be set to a value smaller than “7” and “10”, for example.

  The cluster size number is set as a value representing the assumed maximum size of the community. For example, if the number of cluster sizes is set low, a small community is simulated. If the cluster size is set high, it means that a large community is simulated.

  Further, the threshold value is a value that represents the probability that each source / destination node pair is not connected in the graph data generation process described above. For example, when the threshold value is set low, a community with dense links is simulated, and when the threshold value is set high, a community with thin links is simulated.

  That is, by setting the cluster size number and the threshold value, it is possible to simulate communities having various link densities and sizes. For example, if the threshold value is set low and the cluster size number is set low, a narrow community with dense links is simulated. If the threshold value is set high and the cluster size number is set high, it means that a wide community with few links is simulated.

  Here, in the first and second embodiments, connectivity within the assumed community is not guaranteed as a result of the randomness of link generation. In other words, in the first and second embodiments, in the plurality of graph data generated by the graph data generation unit 12, the number of graph data describing the connectivity in the community may be less than the set cluster size number. is there. The probability that the number of graph data describing connectivity in the community is lower than the cluster size number depends on both the cluster size number and the threshold value. In other words, when the threshold is high, the probability that all nodes do not have mutually reachable paths increases, and as a result, the number of graph data describing connectivity within the community is less than the number of cluster sizes. Probability increases. For this reason, it is desirable to determine the number of cluster sizes in combination with a threshold value.

  Note that if the simulation target is a potential community and sites that tend to be reluctant to link to each other, the threshold value is a test indicating that all nodes are not connected because the links are sparse. It may be a case where it is set higher in order to generate data.

  The test data generated using the setting data set in this way is obtained by superimposing links simulating the margins between nodes belonging to a cluster with respect to the number of cluster divisions that can be attributed to multiple assignments. For example, if it is assumed that a node is a website and a cluster simulates a community to which a website operator belongs, each operator can belong to up to t communities, and even between websites that do not belong directly to the same community. You can simulate how the relationship occurs. Therefore, by using the test data generated in the second embodiment, the operator can verify an important path for determining the strength of relationship between Web sites, for example.

  In the first and second embodiments, the case where the test data generation method is executed by one apparatus has been described. However, the test data generation method described in the first and second embodiments may be executed in a distributed manner by a plurality of apparatuses. For example, in the test data generation method described in the first and second embodiments, the device having the functions of the cluster generation unit 11 and the graph data generation unit 12 and the device having the function of the test data generation unit 13 are one virtual. It may be realized by functioning as the test data generation device 10. In addition, an environment in which a virtual machine is started in a system composed of one device or two or more devices, and a plurality of virtual computers operate on the virtual machine. The test data generation method described in the first and second embodiments may be executed on a computer.

  The test data generation method described in the first and second embodiments can be realized by executing a test data generation program prepared in advance on a computer such as a personal computer or a workstation. In the following, an example of a computer that executes a test data generation program having the same function as that of the test data generation apparatus 10 shown in the first embodiment will be described with reference to FIG. FIG. 13 is a diagram illustrating the computer that executes the test data generation program according to the first embodiment. In the following, a case where the test data generation method is executed by one computer will be described. However, the test data generation method described in the first and second embodiments may be executed in a distributed manner by a plurality of computers as described above. Further, the test data generation method described in the first and second embodiments may be executed on a virtualized computer as described above.

  As shown in FIG. 13, a computer 100 that is an information processing device such as a PC (Personal Computer) has a CPU (Central Processing Unit) 110, an input device 120, an output device 130, a medium reading device 140, and a network interface 150. . The CPU 110 executes various arithmetic processes. The input device 120 receives input of various data from the user. The output device 130 outputs various data processed by the CPU 110. The medium reader 140 reads a program or the like from a storage medium. The network interface 150 exchanges data with other computers via the network. For example, the network interface 150 can receive the setting list via the network.

  The computer 100 also has a RAM 160 and an HDD 170. The RAM 160 temporarily stores various information processed by the CPU 110. The HDD 170 stores various data and various programs.

  For example, the HDD 170 includes a cluster generation program 171, a graph data generation program 172, and a test data generation program having the same functions as the cluster generation unit 11, the graph data generation unit 12, and the test data generation unit 13 illustrated in FIG. 1. 173 is stored. Further, the CPU 110 reads the cluster generation program 171, the graph data generation program 172, and the test data generation program 173 from the HDD 170, and stores the cluster generation program 171, the graph data generation program 172, and the test data generation program 173 in the RAM 160. 161, a graph data generation program 162 and a test data generation program 163.

  Thereafter, the CPU 110 executes the cluster generation program 161, the graph data generation program 162, and the test data generation program 163 as the cluster generation process 111, the graph data generation process 112, and the test data generation process 113, respectively.

  Note that the cluster generation program 171, the graph data generation program 172, and the test data generation program 173 are not necessarily stored in the HDD 170, and are stored in a storage medium such as a CD-ROM and read from the storage medium, for example. May be executed. Alternatively, the cluster generation program 171, the graph data generation program 172, and the test data generation program 173 are stored by another computer and acquired via a public line, the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), or the like. May be executed.

  As described above, the test data generation method, the test data generation device, and the test data generation program according to the present invention are designed to simulate the external input in order to determine the validity of the algorithm used for information processing in which the external input exists. This is useful when generating data, and is particularly suitable for generating test data that simulates a biased link structure within a potential community.

DESCRIPTION OF SYMBOLS 10 Test data generation apparatus 11 Cluster generation part 12 Graph data generation part 13 Test data generation part 20 Input part 30 Output part 40 Input / output control I / F part 50 Storage part 51 Setting list 52 Node list 53 Graph data classified by cluster 54 Test data 60 processor

Claims (7)

Computer
By randomly dividing a node group composed of a plurality of nodes each having identification information into a number of clusters determined by the total number of nodes of the node group and a predetermined number of cluster sizes, a plurality of clusters can be divided. A cluster generation step to generate;
By setting link attribute information indicating a link state between nodes constituting each cluster generated by the cluster generation step using a random number generated between the nodes and a predetermined threshold, A graph data generation step for generating graph data indicating a link structure in each cluster;
All graph data generated by the graph data generation step is assigned to each node by controlling the cluster generation processing by the cluster generation step and the graph data generation processing by the graph data generation step to be repeated a predetermined number of times. A test data generation step of generating test data simulating the link state between the nodes constituting the node group by integrating based on the identified identification information;
The test data generation method characterized by including. The graph data generation step sets, as the link attribute information, a weight indicating the presence or absence of a link between nodes, or a relationship strength between nodes,
In the test data generation step, when the link attribute information is the presence / absence of a link between nodes, the test data is generated by taking the logical sum of all the graph data generated by the graph data generation step, and the link The test data is generated by calculating an arithmetic sum of all graph data generated by the graph data generation step when the attribute information is a weight indicating a relation strength between nodes. Test data generation method.   In the cluster generation step, as a first cluster generation method, clusters are sequentially generated by randomly extracting nodes of the predetermined number of cluster sizes from a node group not extracted as a cluster among the plurality of nodes. Further, when the number of nodes of the node group that has not been extracted as a cluster is smaller than the predetermined cluster size number, the node group that has not been extracted as the cluster is randomly allocated to each of the generated clusters. 3. The test data generation method according to claim 1 or 2, wherein an integer number of clusters that are maximum not more than a value obtained by dividing the total number of nodes of the group by the predetermined number of cluster sizes is generated.   In the cluster generation step, as a second cluster generation method, clusters are sequentially generated by randomly extracting the nodes of the predetermined cluster scale number from a node group not extracted as a cluster among the plurality of nodes. Further, when the number of nodes in the node group that has not been extracted as a cluster is smaller than the predetermined number of cluster sizes, the node group that has not been extracted as the cluster is generated as a new cluster, so that The test data generation method according to claim 1, wherein an integer number of clusters that are minimum at or above a value obtained by dividing a number by the predetermined cluster size number is generated.   The test data generation step includes at least one of the predetermined cluster size number and the predetermined threshold every time the cluster generation process by the cluster generation step and the graph data generation process by the graph data generation step are executed the predetermined number of times. Any one of them is changed, The test data generation method as described in any one of Claims 1-4 characterized by the above-mentioned. By randomly dividing a node group composed of a plurality of nodes each having identification information into a number of clusters determined by the total number of nodes of the node group and a predetermined number of cluster sizes, a plurality of clusters can be divided. A cluster generating means for generating;
By setting link attribute information indicating a link state between nodes constituting each cluster generated by the cluster generation means using a random number generated between the nodes and a predetermined threshold, Graph data generating means for generating graph data indicating the link structure in each cluster;
All graph data generated by the graph data generation unit is assigned to each node by controlling the cluster generation processing by the cluster generation unit and the graph data generation processing by the graph data generation unit to be repeated a predetermined number of times. A test data generating means for generating test data simulating the link state between the nodes constituting the node group by integrating based on the identified identification information;
A test data generating apparatus comprising:   A test data generation program for causing a computer to execute the test data generation method according to claim 1.

Images