Data correspondence estimation device and program

What is AI technical title?
AI technical title is built by PatSnap AI team. It summarizes the technical point description of the patent document.
The data correspondence estimation device addresses the challenge of associating diverse observational data formats by identifying co-occurrence relationships, ensuring accurate object mapping despite data type differences or lack of general relationships, and handling missing or noisy data through optimization.

WO2026126439A1PCT designated stage Publication Date: 2026-06-18NT T INC

View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office: WO · WO
Patent Type: Applications
Current Assignee / Owner: NT T INC
Filing Date: 2024-12-12
Publication Date: 2026-06-18

Smart Images

Figure JP2024044068_18062026_PF_FP_ABST

Patent Text Reader

Abstract

One aspect of the present invention uses a plurality of datasets obtained by sorting a plurality of pieces of first target data and a plurality of pieces of second target data acquired for an object in respective first and second modes that are different from each other according to prescribed units to identify combinations of first target data and second target data that have a co-occurrence relationship between the first mode and the second mode from among all of the combinations of the first target data and the second target data included in the plurality of datasets and associates the identified combinations of the first target data and the second target data with the object.

Need to check novelty before this filing date? Find Prior Art

Description

Data correspondence relationship estimation device and program 【0001】 One aspect of this invention relates to a data correspondence estimation device used to associate multiple observational data recorded in different forms for a common object with an object, and a program used in this device. 【0002】 For example, when managing a communication network, multiple administrators may record observational and additional data on communication equipment, goods, and administrators in their own unique formats. Here, "format" refers to the method of acquiring or recording observational data, such as the names of equipment and goods, the names of administrators, and the data type of the observational data itself. Hereafter, "format" will also be referred to as "mode." 【0003】 If observational data for a common object is recorded in different modes, it may cause problems in terms of efficiency and reliability when managing the communication network. Therefore, it is necessary to associate multiple observational data recorded in different modes with each object. 【0004】 Therefore, conventionally, when multiple observation data recorded in different modes are recorded as the same type of data, such as text data, a method has been proposed in which the similarity between the text data is calculated, and observation data whose evaluation value based on the calculated similarity is equal to or greater than a predetermined value is estimated to correspond to the same object (see, for example, Non-Patent Document 1). 【0005】 Another approach involves, for example, if the administrator's identification information, contained in multiple observation data recorded in different modes, is recorded as either facial image data or audio data, then using a machine learning model that has learned the general correspondence between faces and voiceprints to estimate the correspondence (see, for example, Non-Patent Documents 2 or 3). 【0006】T. Masuyama, S. Sekine, and H. Nakagawa, “Automatic Construction of Japanese KATAKANA Variant List from Large Corpus,” Proceedings of the 20th International Conference on Computational Linguistics (COLING'04), 2:1214-1219. 2004. Tae-Hyun Oh, et al., “Speech2Face: Learning the Face Behind a Voice,” Internet <URL: https: / / arxiv.org / pdf / 1905.09773.> Kei Hashimoto, “Deep Learning-Based Speech Synthesis for Predicting Voice from Face,” Internet <URL: https: / / technofair.web.nitech.ac.jp / wp-content / uploads / 2021 / 10 / TF2021_hashimoto-kei.pdf.> 【0007】 However, the method using Non-Patent Document 1 is applicable when observation data recorded in different modes is described by the same type of data, such as text, but it is not applicable when it is recorded by different types of data, such as text and facial images or voiceprints. Furthermore, the method using Non-Patent Documents 2 or 3 can match data types that have a general relationship, such as the relationship between a face and a voiceprint. However, it is difficult to match data that does not have a general relationship, such as the relationship between the face of a manager and the identification number of the terminal used by that manager. 【0008】 This invention was made in view of the above circumstances, and aims to provide a technology that enables the association of multiple target data recorded in different modes for the same object, even when the data types of the target data are different, or when there is no general relationship between the data. 【0009】To solve the above problems, one aspect of the data correspondence relationship estimation device according to the present invention is configured such that, with respect to an object, a plurality of first target data and a plurality of second target data acquired by different first and second modes, respectively, are classified into predetermined units and a plurality of datasets generated, the device identifies combinations of the first target data and the second target data that are in a co-occurrence relationship between the first mode and the second mode, among all combinations of the first target data and the second target data included in the plurality of datasets, and associates the identified combinations of the first target data and the second target data with the object. 【0010】 According to one aspect of this invention, among all combinations of first target data and second target data acquired by different first and second modes, combinations that have a co-occurrence relationship between the first mode and the second mode are identified, and the identified combinations of first target data and second target data are used as candidates for mapping to an object. Therefore, even if the first and second target data acquired by the first and second modes are recorded by different types of data, or do not have a general relationship with each other, it is possible to identify combinations of target data that correspond to the same object. 【0011】 In other words, according to one aspect of this invention, it is possible to provide a technology that enables the association of multiple target data recorded in different modes for the same object, even when the data types of the target data are different, or when there is no general relationship between the data. 【0012】Figure 1 is a diagram showing an example of a network monitoring system according to the first embodiment of the present invention. Figure 2 is a block diagram showing an example of the hardware configuration of a data correspondence relationship estimation device according to the first embodiment of the present invention. Figure 3 is a block diagram showing an example of the software configuration of a data correspondence relationship estimation device according to the first embodiment of the present invention. Figure 4 is a flowchart showing an example of the processing procedure and processing content of the data correspondence relationship estimation process executed by the control unit of the data correspondence relationship estimation device shown in Figure 3. Figure 5 is a diagram showing the data processing content in the first embodiment of the co-occurrence relationship determination process among the processing procedures shown in Figure 4. Figure 6 is a flowchart showing the second embodiment of the processing procedure and processing content of the co-occurrence relationship determination process among the processing procedures shown in Figure 4. Figure 7 is a flowchart showing the third embodiment of the processing procedure and processing content of the co-occurrence relationship determination process among the processing procedures shown in Figure 4. Figure 8 is a diagram showing the data processing content in the second embodiment of the co-occurrence relationship determination process shown in Figure 6. Figure 9A is a diagram showing the first half of the data processing content in the third embodiment of the co-occurrence relationship determination process shown in Figure 6. Figure 9B is a diagram showing the second half of the data processing content in the third embodiment of the co-occurrence relationship determination process shown in Figure 6. Figure 10 is a block diagram showing an example of the software configuration of a data correspondence relationship estimation device according to a second embodiment of the present invention. Figure 11 is a flowchart showing a first embodiment of the processing procedure and processing content of the observation data classification process executed by the control unit of the data correspondence relationship estimation device shown in Figure 10. Figure 12 is a flowchart showing a second embodiment of the processing procedure and processing content of the observation data classification process executed by the control unit of the data correspondence relationship estimation device shown in Figure 10. Figure 13 is a diagram showing the data processing content in the first embodiment of the observation data classification process shown in Figure 11. Figure 14 is a diagram showing the data processing content in the second embodiment of the observation data classification process shown in Figure 12. Figure 15 is a diagram illustrating an example of observation data by mode. Figure 16 is a diagram illustrating another example of observation data by mode. 【0013】 Embodiments of this invention will be described below with reference to the drawings. 【0014】[First Embodiment] (Configuration Example) (1) System Figure 1 is a diagram showing an example of a network monitoring system according to the first embodiment of the present invention. 【0015】 The network monitoring system according to the first embodiment monitors a network NW that connects a plurality of communication devices SW1 to SW5, including, for example, communication equipment. The network NW is connected to management terminals TM1 to TMn, each used by a plurality of administrators, a plurality of observation data storage devices DB1 to DBn, each consisting of, for example, a memory server, and a data correspondence relationship estimation device SVA. 【0016】 Management terminals TM1 to TMn are used by multiple administrators to input observation data, etc. Observation data storage devices DB1 to DBn each store the observation data input by the management terminals TM1 to TMn. 【0017】 The observation data stored in the observation data storage devices DB1 to DBn is entered by each administrator in their own unique format (mode). For example, the observation data is entered and stored in a format defined by the person managing the data, including the names of communication equipment and items, the names of administrators, and the data types of the observation data. 【0018】 Figure 15 shows an example of observation data defined independently by multiple administrators. In this example, administrators U1 to U3 use their own modes, i.e., independently defined names, for multiple communication equipment as objects, and also input additional information such as the installation location, housing facility, and inspection results of the equipment. For the sake of simplicity, the additional information will be assumed to be included in the observation data from now on. 【0019】 Another example of observation data input in a unique mode is the identification information of the administrator. For example, Figure 16 illustrates a case where the administrator's identification information includes their name, facial photograph, voiceprint, and the terminal identification numbers of the management terminals TM to TMn used by the administrator. 【0020】(2) Data correspondence relationship estimation device SVA The data correspondence relationship estimation device SVA is installed on a server computer located, for example, on the Web or in the cloud. 【0021】 Figures 2 and 3 are block diagrams showing examples of the hardware and software configurations of the data correspondence relationship estimation device SVA, respectively. 【0022】 The data correspondence relationship estimation device SVA includes a control unit 1A having a hardware processor, and a storage unit having a program storage unit 2A and a data storage unit 3, and a communication interface unit (hereinafter referred to as I / F) 4 are connected to this control unit 1A via a bus 5. 【0023】 The communication interface unit 4 transmits and receives information data between the management terminals TM1 to TMn and the observation data storage devices DB1 to DBn, respectively, in accordance with communication protocols such as TCP / IP defined in the network NW. 【0024】 The program storage unit 2A combines, for example, a non-volatile memory that can be written to and read at any time, such as an HDD (Hard Disk Drive) or SSD (Solid State Drive), with a non-volatile memory such as ROM (Read Only Memory), and stores application programs necessary to execute various processes according to the first embodiment of this invention, in addition to middleware such as an OS (Operating System). 【0025】 The data storage unit 3A combines a non-volatile memory that can be written to and read at any time, such as an HDD or SSD, with a volatile memory such as RAM (Random Access Memory). Its storage area includes a data set storage unit 31, a determination data storage unit 32, a co-occurrence relationship information storage unit 33, and a correspondence relationship information storage unit 34. 【0026】The dataset storage unit 31 has multiple datasets pre-stored in it, which are generated by classifying each observation data stored in the respective observation data storage devices DB1 to DBn in different modes into predetermined units, for example, into common time periods of observation. 【0027】 The determination data storage unit 32 stores a list of observed data distribution information that represents the determination result of whether or not observed data is present in each of the above datasets. 【0028】 The co-occurrence relationship information storage unit 33 stores information representing the co-occurrence relationships of observed data between modes, which is determined based on the distribution list information of the observed data. 【0029】 The correspondence relationship information storage unit 34 stores correspondence relationship information that associates each observation data determined to be in a co-occurrence relationship between modes with an object. 【0030】 The control unit 1A includes, as processing functions necessary to realize the first embodiment of this invention, an observation data presence / absence determination processing unit 11, a co-occurrence relationship determination processing unit 12, and an object mapping processing unit 13. 【0031】 Each of the above-mentioned processing units 11 to 13 is implemented by having the hardware processor of the control unit 1A execute an application program stored in the program storage unit 2A. Note that some or all of the above-mentioned processing units 11 to 13 may be implemented using hardware such as LSI (Large Scale Integration) or ASIC (Application Specific Integrated Circuit). 【0032】 The observation data presence / absence determination processing unit 11 determines which observation data is included in each of the multiple datasets stored in the dataset storage unit 31. The observation data presence / absence determination processing unit 11 then generates an observation data distribution list information representing the determination result and stores the generated observation data distribution list information in the determination data storage unit 32. 【0033】The co-occurrence relationship determination processing unit 12 identifies combinations of observed data that have a co-occurrence relationship between modes based on the distribution list information of the observed data, and stores the identified combinations of observed data in the co-occurrence relationship information storage unit 33 as candidates for mapping to an object. 【0034】 The object mapping processing unit 13 maps the combination of observation data stored as mapping candidates in the co-occurrence relationship information storage unit 33 to the corresponding object, and stores information representing that mapping relationship in the mapping relationship information storage unit 34. 【0035】 The processes for determining the presence or absence of the above-mentioned observation data, determining co-occurrence relationships, and mapping to objects will be explained in detail in the operation example. 【0036】 (Example of operation) Next, an example of the operation of the data correspondence relationship estimation device SVA configured as described above will be explained. 【0037】 Figure 4 is a flowchart showing an example of the processing procedure and processing content of the correspondence estimation process performed by the control unit 1A of the data correspondence estimation device SVA. 【0038】 For the purposes of this explanation, we will assume that the observation data set is pre-generated and stored in the data set storage unit 31. 【0039】 (1) Generation of distribution list information of observed data For example, when a correspondence relationship estimation request entered by a manager is detected in step S1, the control unit 1A of the data correspondence relationship estimation device SVA first executes the process of generating distribution list information of observed data under the control of the observation data presence / absence determination processing unit 11 as follows. 【0040】In other words, the observation data presence / absence determination processing unit 11 first reads multiple datasets from the dataset storage unit 31 in step S2. Then, in step S3, the observation data presence / absence determination processing unit 11 determines which observation data is included in each of the read datasets. Based on the determination result, it generates an observation data distribution list information, setting, for example, "1" if observation data is included and "0" if it is not, and stores the generated observation data distribution list information in the determination data storage unit 32. 【0041】 Figure 5 shows an example of the observed data distribution list information. In this example, for each of the datasets DS1, DS2, and DS3, the observed data a １ ~a ５ , b １ ~b ５ , c １ ~c ５ This shows the result of determining whether or not a certain element is present, and displaying the result in a list distributed as "1" or "0". 【0042】 (2) Determination of co-occurrence relationships of observed data The control unit 1A of the data correspondence relationship estimation device SVA then, under the control of the co-occurrence relationship determination processing unit 12, in step S4, executes the process of determining the co-occurrence relationships of the observed data as follows. 【0043】 Several methods can be considered for determining the co-occurrence relationships of observed data. In this embodiment, three determination methods will be described as Example 1, Example 2, and Example 3, respectively. 【0044】 (2-1) Example 1 The co-occurrence relationship determination processing unit 12 reads the observed data distribution list information from the determination data storage unit 32. Based on the read observed data distribution list information, it identifies combinations of observed data that satisfy the determination conditions as data in a co-occurrence relationship. The determination conditions are stored in advance in the condition storage area of the data storage unit 3A. 【0045】First, the conditions that the dataset should satisfy are shown below. (1) The dataset contains observation data for all modes. (2) The observation data for the same mode for the same object always match (it is possible to determine that they match). (3) For any dataset, the observation data is recorded simultaneously for only a part of the plurality of objects targeted. (4) The objects for which the observation data is recorded simultaneously are different depending on the dataset. 【0046】 Note that, among these conditions, for (2), it will be used appropriately depending on how the dataset is prepared, and it is difficult to determine whether it holds in the actual dataset itself. Therefore, the data correspondence relationship estimation device SVA does not perform the determination for (2) above. 【0047】 The co-occurrence relationship determination processing unit 12 examines the presence or absence of each observation data for each mode, and identifies the observation data of different modes whose results match as data that satisfies the co-occurrence relationship. 【0048】 For example, in the example shown in FIG. 5, the "1", "0" distribution patterns of the observation data a １ ~a ５ , b １ ~b ５ , c １ ~c ５ are compared among the modes a, b, c. As a result of this comparison, the observation data with the same "1", "0" distribution pattern is identified as a combination of observation data having a co-occurrence relationship. 【0049】 As a result, in the example of FIG. 5, the observation data a １ , b ３ , c ４ is identified as a combination of observation data having a co-occurrence relationship among the modes a, b, c. Similarly, the observation data a ２ , b ２ , c １ , the observation data a ３ , b ４ , c ５ , the observation data a ４ , b ５, c ２ and observational data a ５ , b １ , c ３ These are identified as combinations of observational data that have co-occurrence relationships between modes a, b, and c, respectively. 【0050】 The co-occurrence relationship determination processing unit 12 stores the combinations of observation data having the co-occurrence relationships identified as described above in the co-occurrence relationship information storage unit 33 as candidates for mapping to an object. 【0051】 (2-2) Example 2 Example 2 relaxes the co-occurrence relationship determination condition (1), and requires that the dataset contains observational data for at least two modes. 【0052】 First, the conditions that the dataset must satisfy in Example 2 are as follows: (1)' The dataset contains observation data for at least two modes. (2) Observation data for the same mode for the same object always matches (it is possible to determine that they match). (3) In each dataset, observation data is recorded simultaneously for only a portion of the multiple objects being targeted. (4) The objects for which observation data is recorded simultaneously differ depending on the dataset. In this case as well, for the same reasons as in Example 1, no determination is made regarding condition (2). 【0053】 Figure 6 is a flowchart showing an example of the processing procedure and processing content of the co-occurrence relationship determination process executed by the co-occurrence relationship determination processing unit 12 in Embodiment 2. 【0054】 In step S41, the co-occurrence relationship determination processing unit 12 first reads the determination conditions to be used in Example 2 from the condition storage area in the data storage unit 3A. Next, in step S42, the co-occurrence relationship determination processing unit 12 reads the observation data distribution list information from the determination data storage unit 32. Then, in step S43, the co-occurrence relationship determination processing unit 12 determines whether the observation data in the observation data distribution list information satisfies the determination conditions, and if it is determined that the conditions are met, in step S44, it identifies the combination of observation data that has a co-occurrence relationship between modes. 【0055】 Figure 8 shows an example of the processing content of the co-occurrence relationship determination process executed by the co-occurrence relationship determination processing unit 12 in Embodiment 2. 【0056】 For example, the co-occurrence relationship determination processing unit 12 determines each observed data a across all datasets DS1 to DS9. １ ~a ５ , b １ ~b ５ , c １ ~c ５ The distribution patterns of "1" and "0" are determined to see if they satisfy the following constraints. The constraints are also stored in advance in the condition storage area within the data storage unit 3A. 【0057】 (5) In each mode, there is exactly one observational data point for the same object; that is, for any two modes, there is exactly one pair of corresponding observational data points. (6) If any two observational data points for any two modes included in the set of observational data points are observational data points for the same object, then the presence or absence of the observational data points will match. 【0058】 Then, the co-occurrence relationship determination processing unit 12 satisfies the above constraints and processes the observed data a of each mode. １ ~a ５ , b １ ~b ５ , c １ ~c ５ For the distribution patterns of "1" and "0" across datasets DS1 to DS3, we solve a combinatorial optimization problem to maximize the number of combinations of observed data. For example, observed data a １ and b ２ , b ２ and c ２ The constraints are met, but a １ and c ２ Since the constraint is not met, a １ , b ２ , c ２ The combination is determined not to have a co-occurrence relationship. 【0059】 As a result, in the example shown in Figure 8, the observed data a １ , b ３ , c ５ , observation data a２ , b ２ , c ２ , observation data a ３ , b ５ , c １ , observation data a ４ , b ４ , c ４ and observational data a ５ , b １ , c ３ Each of these combinations is identified as a combination of observed data that has a co-occurrence relationship between modes a, b, and c. 【0060】 In other words, the co-occurrence relationship determination processing unit 12 identifies a combination of observed data that satisfies the constraints and maximizes the evaluation index as a combination of observed data having a co-occurrence relationship. The co-occurrence relationship determination processing unit 12 then stores the combination of observed data having a co-occurrence relationship identified in this way in the co-occurrence relationship information storage unit 33 as a candidate for mapping to an object. 【0061】 (2-3) Example 3 In Example 3, in addition to condition (1) of the dataset conditions, condition (2) is relaxed so that co-occurrence relationships can be determined even if there are discrepancies due to missing or noisy data in part of the observed data. 【0062】 First, the conditions that the dataset must satisfy in Example 3 are as follows: (1)′ The dataset contains observation data for at least two modes. (2)′ The observation data for the same mode on the same object generally matches, but there are discrepancies due to missing or noisy data. (3) In all datasets, observation data is recorded simultaneously for only some of the multiple objects being targeted. (4) The objects on which observation data is recorded simultaneously differ depending on the dataset. 【0063】 Furthermore, regarding the observation data shown in (2)', it is difficult to determine from the actual dataset itself whether there are any missing or noisy data. Therefore, the data correspondence estimation device SVA does not perform a determination regarding (2) above. 【0064】Figure 7 is a flowchart showing an example of the processing procedure and processing content of the co-occurrence relationship determination process related to Embodiment 3, which is executed by the co-occurrence relationship determination processing unit 12. 【0065】 In step S41, the co-occurrence relationship determination processing unit 12 first reads the determination conditions to be used in Example 3 from the condition storage area of the data storage unit 3A. 【0066】 Next, in step S42, the co-occurrence relationship determination processing unit 12 reads the observation data distribution list information from the determination data storage unit 32. Then, in step S43, the co-occurrence relationship determination processing unit 12 determines whether the observation data in the observation data distribution list information satisfies the determination conditions. If it is determined that the conditions are met, in step S45, it calculates an optimization problem based on the evaluation value, and based on the result, in step S44, it identifies combinations of observation data that have a co-occurrence relationship between modes. 【0067】 Figures 9A and 9B show an example of the processing content of the co-occurrence relationship determination process related to Embodiment 3, which is executed by the co-occurrence relationship determination processing unit 12. 【0068】 In other words, the co-occurrence relationship determination processing unit 12 first examines each observation data a across all datasets DS1 to DS24. １ ~a ５ , b １ ~b ５ , c １ ~c ５ We will determine whether the following constraints are met for the distribution patterns of "1" and "0". 【0069】 (5) In each mode, there is exactly one observational data point for the same object, meaning that for any two modes, there is exactly one combination of observational data points that can be associated. 【0070】 Then, the co-occurrence relationship determination processing unit 12 satisfies the above constraints and uses the observed data a for each mode a, b, c. １ ~a ５ , b １ ~b ５ , c １ ~c ５We solve a combinatorial optimization problem for the distribution patterns of "1" and "0" such that the evaluation value, which increases monotonically with respect to the number or proportion of times the observed data matches, is maximized. 【0071】 For example, in the example shown in Figure 9A, the observed data b in dataset DS4 ３ , observation data c in dataset DS12 １ , and observational data a in dataset DS21 ２ This shows a case where observational data that should actually exist is considered "none (0)" due to, for example, missing data or noise. 【0072】 Figure 9B shows the observed data for each mode a, b, and c. １ ~a ５ , b １ ~b ５ , c １ ~c ５ This shows the relationship between the two modes, where the denominator represents the number of times the dataset contains observational data for both modes, and the numerator represents the number of times the presence or absence of observational data for both modes matches. 【0073】 The co-occurrence relationship determination processing unit 12 generates multiple correspondence patterns by combining the observation data shown in Figure 9B between modes a, b, and c. For example, Figure 9B shows the case where correspondence pattern K1 is generated by combining the observation data enclosed in circles indicated by dashed lines, and correspondence pattern K2 is generated by combining the observation data enclosed in circles indicated by solid lines. 【0074】 The co-occurrence relationship determination processing unit 12 calculates an evaluation value for each of the generated correspondence patterns K1 and K2. For example, it calculates the sum of the number of times the presence or absence of observed data matches (evaluation value 1) and the product of the proportion of times the presence or absence of observed data matches (evaluation value 2). 【0075】 First, regarding the correspondence pattern K1, the evaluation value 1 is calculated as follows: That is, observation data a １ , b ３ Evaluation value for the combination and observation data b ３ , c ５ Evaluation value for the combination and observation data c ５ , a１ Find the sum (7 + 8 + 8) of the evaluation values for the combination. Similarly, for the observed data a ２ , b ２ , c ２ , find the sum (8 + 8 + 7) of the evaluation values for the combination, and for the observed data a ３ , b ５ , c １ , find the sum (8 + 7 + 8) of the evaluation values for the combination, and for the observed data a ４ , b ４ , c ４ , find the sum (8 + 8 + 8) of the evaluation values for the combination, and for the observed data a ５ , b １ , c ３ , find the sum (8 + 8 + 8) of the evaluation values for the combination. Then, find the total sum of the obtained sums of the evaluation values. As a result, evaluation value 1 is calculated to be 117. 【0076】 Also, evaluation value 2 is obtained as follows. That is, find the product (7 / 8 × 8 / 8 × 8 / 8) of the ratios of the combinations of observed data a １ , b ３ , the ratios of the combinations of observed data b ３ , c ５ , and the ratios of the combinations of observed data c ５ , a １ . Similarly, find the product (8 / 8 × 8 / 8 × 7 / 8) of the ratios of the combination of observed data a ２ , b ２ , c ２ , the product (8 / 8 × 7 / 8 × 8 / 8) of the ratios of the combination of observed data a ３ , b ５ , c １ , the product (8 / 8 × 8 / 8 × 8 / 8) of the ratios of the combination of observed data a ４ , b ４ , c ４ , and the product (8 / 8 × 8 / 8 × 8 / 8) of the ratios of the combination of observed data a ５ , b １ , c ３ . Then, calculate the product of the obtained products. As a result, evaluation value 2 is calculated to be (7 / 8) ３ = 0.67. 【0077】On the other hand, for correspondence pattern K2, evaluation value 1 is calculated as follows: that is, observation data a １ , b ２ , c ３ The sum of the evaluation values for the combination (4 + 5 + 4) and the observed data a ２ , b ３ , c ４ The sum of the evaluation values for the combination (5 + 6 + 3) and the observed data a ３ , b ４ , c ５ The sum of the evaluation values for the combination (3 + 6 + 5) and the observed data a ４ , b １ , c １ The sum of the evaluation values for the combination (3 + 6 + 2) and the observed data a ５ , b ５ , c ２ The sum of the evaluation values (4 + 2 + 3) for each combination is calculated, and the sum of these individual sums is then calculated. As a result, the evaluation value for 1 is 61. 【0078】 Furthermore, evaluation value 2 is calculated as follows: that is, observation data a １ , b ２ , c ３ The product of the proportions of the combinations (4 / 8 × 5 / 8 × 4 / 8) and the observed data a ２ , b ３ , c ４ The product of the proportions of the combinations (5 / 8 × 6 / 8 × 3 / 8) and the observed data a ３ , b ４ , c ５ The product of the proportions of the combinations (5 / 8 × 6 / 8 × 3 / 8) and the observed data a ４ , b １ , c １ The product of the proportions of the combinations (3 / 8 × 6 / 8 × 2 / 8) and the observed data a ５ , b ５ , c ２ The product of the proportions of each combination (4 / 8 × 2 / 8 × 3 / 8) is calculated. Then, the product of these products is calculated. As a result, the evaluation value 2 = 0.000016 is calculated. 【0079】The co-occurrence relationship determination processing unit 12 compares one of the evaluation values 1 or 2 calculated for each of the correspondence patterns K1 and K2 as described above, and selects the correspondence pattern with the higher evaluation value. Then, it stores the combination of observation data constituting the selected correspondence pattern in the co-occurrence relationship information storage unit 33 as a candidate for correspondence to the object. 【0080】 In the above explanation, we used the example of creating two correspondence patterns K1 and K2, comparing their evaluation values, and selecting the correspondence pattern with the higher evaluation value. However, in reality, evaluation values are calculated for all possible correspondence patterns, and they are compared with each other. The correspondence pattern with the highest evaluation value is then selected. 【0081】 (3) Finally, in step S5, the control unit 1A of the object correspondence relationship estimation device SVA reads the candidate correspondences of the observed data from the co-occurrence relationship information storage unit 33 under the control of the object correspondence processing unit 13, and associates the read candidate correspondences of the observed data with the corresponding object. The object correspondence processing unit 13 then stores the information representing the correspondence relationship between the object and the observed data in the correspondence relationship information storage unit 34. 【0082】 Furthermore, when the control unit 1A of the data correspondence relationship estimation device SVA receives a transmission request from, for example, a management person, in step S7, it reads information representing the correspondence relationship between the object and the observed data from the correspondence relationship information storage unit 34 and transmits it from the communication I / F unit 4 to the management terminals TM1 to TMn used by the requesting management person. 【0083】 (Effect) As described above, in the first embodiment, observation data a stored in different modes a, b, and c １ ~a ５ , b １ ~b ５ , c １ ~c ５For the multiple datasets DS1 to DS3 generated by classifying them according to the observed time period, the first step is to determine which observation data is included in each dataset and generate an observation data distribution list information showing the determination result. Next, based on the observation data distribution list information above, combinations of observation data that have a co-occurrence relationship between modes a, b, and c are identified, and the identified combinations of observation data are associated with objects. 【0084】 Therefore, even if observational data recorded in different modes are recorded by different types of data, or even if observational data recorded in different modes do not have a general relationship, it is possible to estimate combinations of observational data corresponding to the same object by integrating observational data recorded at multiple time periods and determining the co-occurrence relationships between the modes of that combination. 【0085】 Furthermore, when determining co-occurrence relationships, the determination criteria are partially relaxed, and the combination optimization problem is solved to maximize the number of corresponding observed data. This allows for the identification of combinations of observed data that co-occur among modes a, b, and c. Therefore, even if the dataset does not contain all observed data for modes a, b, and c, combinations of observed data that co-occur can be identified as candidates for mapping to an object. 【0086】 Furthermore, when determining co-occurrence relationships, the system solves a combinatorial optimization problem to maximize an evaluation value that monotonically increases with respect to the number of times or the proportion of times the presence or absence of observed data matches, thereby determining combinations of observed data that have a co-occurrence relationship. Therefore, even if some observed data could not be obtained due to data loss or noise, combinations of observed data that have a co-occurrence relationship between modes a, b, and c can be identified as candidates for mapping to an object. 【0087】[Second Embodiment] In the second embodiment of this invention, the control unit of the data correspondence relationship estimation device is further equipped with an observation data classification processing function. This observation data classification processing function acquires observation data groups recorded in different modes in the observation data storage devices DB1 to DBn on the system, and generates multiple datasets by classifying each acquired observation data group according to predefined classification conditions. 【0088】 (Configuration Example) Figure 10 is a block diagram showing an example of the software configuration of the data correspondence relationship estimation device SVB according to the second embodiment of the present invention. In this figure, the same parts as in Figure 3 are denoted by the same reference numerals, and detailed explanations are omitted. 【0089】 In addition to the observation data presence / absence determination processing unit 11, the co-occurrence relationship determination processing unit 12, and the object mapping processing unit 13, the control unit 1B further includes an observation data classification processing unit 14. 【0090】 Furthermore, this observation data classification processing unit 14, like the observation data presence / absence determination processing unit 11, the co-occurrence relationship determination processing unit 12, and the object mapping processing unit 13, is realized by having the hardware processor of the control unit 1B execute the application program stored in the program storage unit 2B. 【0091】 The observation data classification processing unit 14 acquires observation data sets from observation data storage devices DB1 to DBn, and classifies each acquired observation data set according to classification conditions that define predetermined units, such as the date and time of observation or the content of the observation data, to generate multiple datasets. The observation data classification processing unit 14 then stores the generated multiple datasets in the dataset storage unit 31. 【0092】 (Example of operation) (1) The control unit 1B of the observation data classification data correspondence estimation device SVB performs the following process of classifying each observation data based on pre-set classification conditions under the control of the observation data classification processing unit 14. 【0093】Several methods can be considered for classifying the observed data. In the second embodiment, two classification methods will be described as Example 1 and Example 2. 【0094】 (1-1) Example 1 Observation data storage devices DB1, DB2, and DB3 each store observation data entered by multiple administrators using different, unique modes. 【0095】 For example, as shown in Figure 10, the observation data storage device DB1 contains management data for the building entry application system, that is, data that associates the building name with information such as the time when entry is permitted, and observation data a １ ~a ５ It is remembered as such. 【0096】 Furthermore, the observation data storage device DB2 contains management data for the construction arrangement system, that is, data that associates the name of the construction building with information representing the responsible construction company, etc., and observation data b １ ~b ５ It is remembered as such. 【0097】 Furthermore, the observation data storage device DB3 contains management data for the communication equipment management system, that is, data that associates the communication building name with information representing the accommodation area of each piece of equipment, and observation data c １ ~c ５ It is remembered as such. 【0098】 The control unit 1B of the data correspondence relationship estimation device SVB performs the observation data classification process related to Embodiment 1 as follows, under the control of the observation data classification processing unit 14. 【0099】 Figure 11 is a flowchart showing an example of the processing procedure and processing content of the observation data classification process according to Embodiment 1, which is executed by the observation data classification processing unit 14. 【0100】 In step S81, the observation data classification processing unit 14 first accesses each of the observation data storage devices DB1 to DB3, and then extracts observation data a from each of these observation data storage devices DB1 to DB3. １ ~a ５ , b １ ~b ５ , c １ ~c ５The data is acquired via the communication I / F unit 4. Then, each acquired observation data is a １ ~a ５ , b １ ~b ５ , c １ ~c ５ This is stored in the observation data storage area (not shown) within the data storage unit 3B. 【0101】 Next, in step S82, the observation data classification processing unit 14 reads the classification conditions used in Example 1 from the condition storage area of the data storage unit 3B. Then, in step S83, the observation data classification processing unit 14 reads each observation data a from the observation data storage area in the data storage unit 3B. １ ~a ５ , b １ ~b ５ , c １ ~c ５ Each of these is read, and the above observation data a is read. １ ~a ５ , b １ ~b ５ , c １ ~c ５ The process of classifying the data according to the above classification criteria is executed. 【0102】 For example, if the entry date, construction date, and installation date are defined as classification conditions, the observation data classification processing unit 14 will classify the observation data a related to the entry application. １ ~a ５ The entries are classified by the date of entry. As a result, as shown in Figure 13, for example, the entry applications are classified in a way that corresponds to the entry dates "9 / 2", "9 / 5", and "9 / 17". Note that the above entry applications are not the observation data itself, but rather additional data that shows the usage status of the observation data. However, for the sake of explanation, they are treated as part of the observation data here. 【0103】 Furthermore, the observation data classification processing unit 14 processes the observation data b related to the above construction arrangements. １ ~b ５ The requests are then classified by construction date. As a result, as shown in Figure 13, for example, the construction requests are classified in a way that they correspond to the construction dates "9 / 2", "9 / 5", and "9 / 17". 【0104】Similarly, the observation data classification processing unit 14 processes the observation data c related to the management of the communication equipment. １ ~c ５ The reports are classified according to the installation date of the communication equipment. As a result, as shown in Figure 13, for example, the communication equipment change reports are classified in a way that corresponds to the respective construction dates of "9 / 2", "9 / 5", and "9 / 17". 【0105】 The observation data classification processing unit 14 then generates a dataset containing observation data corresponding to each of the classified modes a, b, and c, for each date, and stores each of the generated datasets in the dataset storage unit 31. 【0106】 For example, in the example shown in Figure 13, observation data a indicating the building name is for the data collected on September 2nd. １ , a ２ , a ３ Observation data b showing the name of the construction building ２ , b ３ , b ４ Observation data c showing the name of the communications building. １ , c ４ , c ５ Dataset DS1 containing this data is generated. 【0107】 Furthermore, for the data collected on September 5th, observation data a indicating the building name is available. ２ , a ３ , a ４ Observation data b showing the name of the construction building ２ , b ４ , b ５ Observation data c showing the name of the communications building. １ , c ２ , c ５ A dataset DS2 containing this data is generated. 【0108】 Similarly, for the data collected on September 17th, observation data a indicating the building name is available. ２ , a ５ Observation data b showing the name of the construction building １ , b ２ Observation data c showing the name of the communications building. １ , c ３ A dataset DS3 containing this data is generated. 【0109】(1-2) Example 2 In Example 2, when classifying observation data, the observation data for each mode obtained from the observation data storage devices DB1 to DB3 is classified using a third data that is related to information representing the usage status of these observation data as a classification criterion. 【0110】 The observation data storage device DB4 stores, for example, management data from a project management system. Figure 14 shows an example. In this example, the management data from the project management system stores information such as a budget management ID associated with the project name. 【0111】 Figure 12 is a flowchart showing an example of the processing procedure and processing content of the observation data classification process according to Embodiment 2, which is executed by the observation data classification processing unit 14. In this figure, the same reference numerals are used for parts that are the same as in Figure 11, and detailed explanations are omitted. 【0112】 In step S81, the observation data classification processing unit 14 first accesses, for example, each observation data storage device DB2 and DB3, and then extracts observation data b from these observation data storage devices DB2 and DB3. １ ~b ５ , c １ ~c ５ The data is acquired via the communication I / F unit 4. Then, each acquired observation data is b １ ~b ５ , c １ ~c ５ This is stored in the observation data storage area within the data storage unit 3B. 【0113】 Furthermore, in step S85, the observation data classification processing unit 14 accesses the observation data storage device DB4 and obtains management data from the project management system via the communication I / F unit 4. The obtained management data is then stored in the condition storage area within the data storage unit 3B. 【0114】 As a result, in the example shown in Figure 14, for example, the information that associates the name of the construction building in the construction scheduling system with the name of the construction company in charge is observed data b. １ ~b ５Furthermore, the information that associates the name of the communication building in the communication equipment management system with information representing the accommodation area is observed data c. １ ~c ５ These are each acquired as such, and further information that associates the budget management ID with the project name in the project management system is acquired as information representing the classification criteria. 【0115】 Next, in step S82, the observation data classification processing unit 14 reads the classification conditions to be used in Example 2 from the condition storage area of the data storage unit 3B. Then, in step S86, the observation data classification processing unit 14 reads each observation data b from the observation data storage area in the data storage unit 3B. １ ~b ５ , c １ ~c ５ Each of these is read, and the above observation data b is read. １ ~b ５ , c １ ~c ５ The process of classifying the data according to the above classification criteria is executed. 【0116】 For example, the observation data classification processing unit 14 processes the observation data b １ ~b ５ These are classified into those with the same billing name, and observation data c １ ~c ５ The data is then classified into groups based on the project name. The observation data classification processing unit 14 then integrates the classified observation data based on the corresponding project name and budget management ID to generate a dataset, and stores each of the generated datasets in the dataset storage unit 31. 【0117】 As a result, as shown in Figure 14, observation data b ２ , b ３ , b ４ And observation data c １ , c ４ , c ５ Dataset DS1 including and observation data b ２ , b ４ , b ５ And observation data c １ , c ２ , c ５ Dataset DS2, which includes and observational data b１ , b ２ And observation data c １ , c ３ A dataset DS3 containing the above is generated. 【0118】 (Effects) As described above, in the second embodiment of this invention, the control unit 1B is further equipped with an observation data classification processing unit 14, and the observation data classification processing unit sorts observation data a recorded in different modes a, b, and c from the observation data storage devices DB1 to DBn on the system. １ ~a ５ , b １ ~b ５ , c １ ~c ５ Each of the acquired observation data a is obtained. １ ~a ５ , b １ ~b ５ , c １ ~c ５ The system generates a dataset by classifying the data according to predetermined units, such as the date and time of observation and pre-defined classification conditions that define the content of the observed data. 【0119】 Therefore, within the data correspondence estimation device SVB, it becomes possible to classify observation data recorded in different modes according to arbitrary classification conditions specified by, for example, the administrator, and estimate the correspondence between the observation data. 【0120】 [Other Embodiments] (1) In the co-occurrence relationship determination processing unit 12, if multiple correspondence candidates are obtained for the same object, and if it is possible to calculate the similarity between observation data of different modes or to estimate using a machine learning model, the calculation of the similarity or estimation using a machine learning model may be performed for each of the multiple correspondence candidates. In this way, it is possible to make correspondences to objects with even greater accuracy. 【0121】(2) In the first and second embodiments, the function of the data correspondence relationship estimation device was described using the example of providing the function on a server computer on the Web or in the cloud. However, the invention is not limited to these cases, and for example, the function of the data correspondence relationship estimation device may be provided on a personal computer used by a system administrator. Alternatively, the function of the data correspondence relationship estimation device may be distributed across multiple server computers or personal computers. 【0122】 (3) In addition, the functional configuration of the data correspondence relationship estimation device, the processing procedure and content of the data correspondence relationship estimation process, the types of data to be associated, the classification conditions of the target data, etc., can be modified in various ways without departing from the spirit of this invention. 【0123】 Although embodiments of this invention have been described in detail above, the above description is merely illustrative in all respects. It goes without saying that various improvements and modifications can be made without departing from the scope of this invention. In other words, when implementing this invention, specific configurations may be adopted as appropriate depending on the embodiment. 【0124】 In short, this invention is not limited to the embodiments described above, and in the implementation stage, the components can be modified and materialized without departing from the gist of the invention. Furthermore, various inventions can be formed by appropriately combining the multiple components disclosed in each of the embodiments. For example, some components may be deleted from all the components shown in each embodiment. Moreover, components from different embodiments may be appropriately combined. 【0125】 SVA, SVB...Data correspondence relationship estimation device DB1 to DBn...Observation data storage device TM to TMn...Management terminal NW...Network 1A, 1B...Control unit 2A, 2B...Program storage unit 3A, 3B...Data storage unit 4...Communication I / F unit 5...Bus 11...Observation data presence / absence determination processing unit 12...Co-occurrence relationship determination processing unit 13...Object mapping processing unit 14...Observation data classification processing unit 31...Data set storage unit 32...Determination data storage unit 33...Co-occurrence relationship information storage unit 34...Correspondence relationship information storage unit

Claims

1. A data correspondence relationship estimation device comprising: a storage medium for storing a plurality of datasets obtained by classifying a plurality of first target data and a plurality of second target data acquired by different first and second modes for an object into predetermined units; a first processing unit that identifies, based on the plurality of datasets stored in the storage medium, combinations of the first target data and the second target data that have a co-occurrence relationship between the first mode and the second mode from among all combinations of the first target data and the second target data included in the plurality of datasets; and a second processing unit that associates the identified combinations of the first target data and the second target data with the object.

2. The data correspondence relationship estimation device according to claim 1, further comprising a third processing unit that determines which target data is included in each of the plurality of datasets and generates target data distribution information representing the determination result, wherein the first processing unit identifies combinations of first target data and second target data that are in a co-occurrence relationship between the first mode and the second mode based on the target data distribution information.

3. The data correspondence relationship estimation device according to claim 1, further comprising a fourth processing unit that acquires a plurality of first target data stored in the first mode and a plurality of second target data stored in the second mode for the object, and classifies the acquired plurality of first target data and plurality of second target data into a plurality of data groups according to classification conditions representing a predetermined unit to generate a plurality of datasets.

4. A program that causes a processor in the data correspondence relationship estimation device to execute at least one of the processes performed by each processing unit in the data correspondence relationship estimation device described in claim 1 or 2.