Supply chain network creation method, supply chain network creation device, and program

Abstract

In this supply chain network creation method, which is performed by a computer system, the computer system includes an arithmetic device and a storage device, and the storage device holds a learning model and transaction history information. The supply chain network creation method includes: a procedure in which the arithmetic device uses the learning model on the basis of input information relating to an article of interest to estimate article items that constitute the article; a procedure in which the arithmetic device converts each of character strings of the estimated items and character strings of items included in the transaction history information into names defined by predetermined rules; and a procedure in which the arithmetic device compares items estimated on the basis of the converted names with the items included in the transaction history information to estimate a supplier company of the articles of the estimated items.

Description

Supply chain network creation method, supply chain network creation device, and program Incorporation by Reference

[0001] This application claims priority from Japanese Patent Application No. 2024-084664, filed on May 24, 2024, the contents of which are incorporated herein by reference.

[0002] The present invention relates to a technique for creating a supply chain network.

[0003] Recently, the manufacturing and distribution industry has been facing increasing risks that threaten stable production and business continuity, such as shortages of semiconductors and other materials, and rising raw material and energy prices, triggered by natural disasters, financial crises, terrorism, and pandemics. Furthermore, as businesses expand globally, the companies that make up the supply chain for a single product—suppliers, manufacturers, wholesalers, logistics companies, and distributors—are increasingly dispersed around the world, increasing exposure to these risks. In response to these circumstances, attention is being focused on strengthening supply chains (SCs). Specifically, companies must understand the companies that make up their SCs, identify risks that could affect production activities, and take proactive measures. They must also be able to quickly grasp the situation and take immediate action in the event of an incident. However, while companies are aware of their direct suppliers (referred to as Tier 1 for convenience) and the products they deal with, information on suppliers involved in the production of each Tier 1 product (referred to as Tier 2 for convenience) is confidential, making it difficult for them to understand the companies and products upstream of Tier 1. Currently, investigations must be conducted using a large amount of manpower or estimation and identification must be performed using some other method. As a method for identifying a supply chain in which multiple companies are hierarchically connected, Japanese Patent Laid-Open Publication No. 2023-175558 (Patent Document 1) describes an information processing system including: a supply chain network acquisition unit that acquires a supply chain network in which multiple nodes corresponding to multiple companies are connected by links indicating business relationships that associate product provider companies with product recipient companies; a subnetwork extraction unit that extracts from the supply chain network a subnetwork that includes at least one of an upstream subnetwork including an upstream company that supplies products to a first company among the multiple companies and a downstream subnetwork including downstream companies that receive products from the first company; a route extraction unit that extracts a group of routes in the subnetwork, the group of routes including one or more routes in which a node representing the first company is connected to a node representing a second company different from the first company by one or more links; and a redrawing unit that redraws the group of routes extracted by the route extraction unit.When selecting a company upstream of Tier 1, the following method is described: "The second company may be specified, for example, by a user performing a predetermined operational input on the input receiving unit 114. Alternatively, the route extraction unit 113 may automatically determine as the second company a company that meets the specified conditions based on the conditions specified and received by the input receiving unit 114. The second company may be, for example, a regulated company whose transactions are restricted, or more specifically, may be a company that is problematic from the perspective of ESG (Environment, Social, Governance). Alternatively, the second company may be a company that corresponds to a choke point in the sub-network 122."

[0004] Patent Document 1: Japanese Patent Application Laid-Open No. 2023-175558

[0005] In Patent Document 1, a supply chain network is represented in a format where nodes indicate companies and links that associate product ordering companies with order receiving companies are connected, but in order to obtain a network related to a specific product, the user needs to specify some conditions, and more specifically, needs to have some knowledge of information about companies and items upstream of Tier 1. However, as mentioned above, it is estimated that there are not many users who have knowledge of information about companies and items upstream of Tier 1.

[0006] The present invention has been made in consideration of these points, and aims to create information on the parts configuration (BOM: Bill of Materials) of products traded between an ordering company (Tier N-1) and an order-receiving company (Tier N), and to hierarchically extract companies (Tier N+1) that have business relationships with Tier N, from N=1 to n.

[0007] In order to solve at least one of the above problems, the present invention provides a supply chain network creation method using a computer system, the computer system comprising an arithmetic unit and a storage device, the storage device holding a learning model and transaction history information, the learning model having been trained to output information indicating item names of items that make up a target item in response to input of information about the target item, the transaction history information including information correlating item names with a company that has ordered the item and a company that has received an order for the item, the supply chain network creation method comprising the steps of: a first step in which the arithmetic unit uses the learning model to estimate item names of items that make up the target item based on the input information about the target item; a second step in which the arithmetic unit converts character strings of the estimated item and character strings of items included in the transaction history information into names defined according to predetermined rules; and a third step in which the arithmetic unit compares the estimated item with items included in the transaction history information based on the converted names to estimate a supplier company of the estimated item.

[0008] One aspect of the present invention allows for the creation of a supply chain network associated with the production of a product, all the way up to upstream suppliers.

[0009] Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiments.

[0010] 1 is a block diagram showing an example of the configuration of a supply chain network creation device according to the present embodiment. FIG. 2 is a flowchart showing an example of the operation of a parts structure generation unit according to the present embodiment. FIG. 3 is an explanatory diagram showing an example of item information read by the parts structure generation unit according to the present embodiment. FIG. 4 is an explanatory diagram showing an example of an output result from a learning model according to the present embodiment. FIG. 5 is an explanatory diagram showing an example of BOM information output by the parts structure generation unit according to the present example. FIG. 6 is a flowchart showing an example of the operation of a trading company inference unit according to the present embodiment. FIG. 7 is an explanatory diagram showing an example of a BOM information image based on BOM information according to the present embodiment. FIG. 8 is an explanatory diagram showing an example of transaction history information held by a supply chain network creation device according to the present embodiment. FIG. 9 is an explanatory diagram showing an example of product code information referenced by a trading company inference unit according to the present embodiment. FIG. 10 is a flowchart showing an example of standard name conversion processing by a trading company inference unit according to the present embodiment. FIG. 11 is an explanatory diagram showing an example of a process by which a trading company inference unit according to the present embodiment converts an input item name into a standard name. FIG. 12 is an explanatory diagram showing an example of SC network information held by a supply chain network creation device according to the present embodiment. FIG. 13 is an explanatory diagram showing an example of a supply chain network display screen output by a supply chain network creation device according to the present embodiment. FIG. 14 is a block diagram showing an example of the hardware configuration of a supply chain network creation device according to the present embodiment.

[0011] A supply chain network creation device according to an embodiment of the present invention will be described below with reference to the drawings.

[0012] FIG. 1 is a block diagram showing an example of the configuration of a supply chain network creation device according to this embodiment.

[0013] The supply chain network creation device 100 comprises an input unit 110, a calculation unit 120, a memory unit 130 that stores input / output information, a communication unit 140, an output unit 150, and a display unit 160. The memory unit 130 has at least a learning model 131, transaction history information 132, BOM information 133 that stores a BOM (Bill of Materials) that is the component configuration of a product, and SC network information 134 that stores information about the supply chain. The input unit 110 may be connected to a communication means 200 such as a network.

[0014] In this embodiment, "products" and "parts" are described separately, but this is done for the sake of convenience in description of distributed goods. In reality, "products" may be treated as "parts," and "parts" may be treated as "products." For example, a "product" shipped by a company on a supply chain network may be incorporated as a "part" into a product of a downstream company. In other words, in the description of this embodiment, "parts" may sometimes be replaced with "products," and "products" may sometimes be replaced with "parts." The term "goods" may be used to encompass these terms.

[0015] The operation of the supply chain network creation device 100 includes a parts structure generation unit 121 and a trading company inference unit 122 in the calculation unit 120. Hereinafter, the operation of the parts structure generation unit 121 and the trading company inference unit 122 will be described using the supply chain network creation device 100 as an example.

[0016] <Parts Structure Generation> FIG. 2 is a flowchart showing an example of the operation of the parts structure generation unit 121 according to this embodiment.

[0017] The parts configuration generation process performed by the supply chain network creation device 100 in this embodiment operates when a processing command is detected from the input unit 110, and generates BOM information using a learning model. A learning model refers to a mechanism for deriving results from input data, and examples of this include machine learning models and generation AI. In this embodiment, the operation will be explained using generation AI as an example.

[0018] The parts configuration generation processing command includes the name of an item for which a supply chain network is to be created. In step S201, the parts configuration generation unit 121 receives the item name and reads information about the item as needed.

[0019] FIG. 3 is an explanatory diagram showing an example of item information read by the parts structure generation unit 121 according to this embodiment.

[0020] 3 is attribute information about an item, and includes an item name 301, a model name 302, a use 303, and a price 304. The item information 300 may be stored in advance in the storage unit 130, or may be received via the input unit 110 in step S201. For example, for a "camera," a message indicating that input of the model name and use is being received is displayed on the display unit 160, and the input is received. In the example of the item information 300 shown in FIG. 3, camera, lens unit, motor, etc. are stored as the item name 301, and information such as the model name 302, use 303 (for example, industrial use), and price 304 for each is stored.

[0021] In step S202, the parts structure generation unit 121 creates a character string to be input to the learning model 131. The character string includes the item name and item information read in step S201, and at least one character string is created. For example, when creating a BOM for an item name of "camera," the character string would be "Please tell me the parts for this camera." Furthermore, if the item information illustrated in FIG. 3 is used, possible character strings would be, for example, "Please tell me the parts for an industrial camera," "Please tell me the parts for a camera with a model name of CA001," or "Please tell me the parts for an industrial camera that cost approximately 100,000 yen."

[0022] In step S203, the parts structure generation unit 121 inputs the created character string to the learning model 131 and extracts the item name of the part from the character string that is the output result.

[0023] FIG. 4 is an explanatory diagram showing an example of an output result from the learning model 131 according to this embodiment.

[0024] In this example, the names of the parts are output in bullet points, and the string from after "number." to before ":" is extracted. The example in Figure 4 shows an example where, when a string such as "Please tell me the parts of a lens unit" is input to the learning model 131, information about the parts that make up a lens unit, such as the lens, AF motor, and aperture, is output. For example, "lens" is extracted from "1. Lens: This changes the direction of light and creates an image of the subject. There are various types, such as single-focus lenses and zoom lenses."

[0025] In order to output the data as described above, the learning data used in advance to create the learning model 131 must include at least information about the components that make up the target product (a camera in the above example). Specifically, the learning data may be based on research papers and articles in specialized journals related to the target product, or publicly available design information may be learned. Non-public information may also be used if it is available for learning, for example, with permission.

[0026] In step S204, the parts structure generation unit 121 checks whether one or more item names that become parts have been extracted, and if so, outputs them to the BOM information 133 in step S205.

[0027] FIG. 5 is an explanatory diagram showing an example of the BOM information 133 output by the parts structure generation unit 121 according to this embodiment.

[0028] The BOM information 133 shown in FIG. 5 is a registration of the parent-child relationships of items, and at least a parent item name 501 and a child item name 502 that is a part of the parent item are registered.

[0029] To generate a BOM for the part registered in step S205, in step S206, the parts structure generation unit 121 adds the part to a list of items for which BOM generation is to be performed. The list is stored in the memory of the calculation unit 120 and is retained until the parts structure generation process is completed. Furthermore, in step S206, the parts structure generation unit 121 retrieves the next item for which BOM generation is to be performed, for example, in the order in which it is stored in the list, and passes information about that item to step S202. In this way, BOM generation is repeated sequentially for the parts found in step S203, and the process ends when the number of items for which BOM generation is to be performed stored in the list in step S206 reaches zero.

[0030] For example, if the user of the supply chain network creation device 100 in this embodiment is a camera manufacturer, the user knows the Tier 1 parts, such as lens units, that the user directly deals with suppliers for camera manufacturing. The parts structure generation unit 121 then creates a string inquiring about the parts that make up the lens unit (step S202) and acquires information about items that make up the lens unit, such as the "lens," "AF motor," and "aperture," from the output of the learning model 131 shown in FIG. 4 (step S203). The parts structure generation unit 121 then outputs these items to the BOM information 133 (step S205) and adds them to a list of items for which BOM generation is required (step S206).

[0031] Next, the parts structure generation unit 121 selects one of the acquired component items of the lens unit, for example, "lens," creates a character string inquiring about the components that make up the lens (step S202), and inputs the character string into the learning model 131 (step S203). As a result, if the components that make up the lens are found (step S204: Yes), the parts structure generation unit 121 outputs those items to the BOM information 133 (step S205), and further adds them to a list of items for which BOM generation is to be performed (step S206).

[0032] On the other hand, if no parts that make up the lens are found (step S204: No), the parts structure generation unit 121 deletes "lens" from the list of items to be generated in the BOM (step S206), selects another of the lens unit's components, for example, "AF motor," creates a string inquiring about the parts that make up the AF motor (step S202), and inputs this string into the learning model 131 (step S203).

[0033] The same process is repeated until all items that make up the lens unit have been deleted from the list of items for which a BOM is to be generated. Then, Tier 1 parts that make up the camera (if any) other than the lens unit are selected, and the same process as above is repeated. This generates a BOM for the target product.

[0034] <Transaction Company Inference> FIG. 6 is a flowchart showing an example of the operation of the transaction company inference unit 122 according to this embodiment.

[0035] The trading company inference process performed by the supply chain network creation device 100 in this embodiment operates when a processing command is detected from the input unit 110, and matches the item names in the BOM information with the item names in the transaction history information, and infers the trading relationships between the companies from the connections between the items in the BOM information.

[0036] The processing command for estimating trading companies includes the company and item name for which a supply chain network is to be created. To facilitate understanding, in this embodiment, the processing will be explained using an example in which a supply chain network for Company X1 and Camera is created when Company: Company X1 and Item: Camera are specified.

[0037] In step S601, the trading company inference unit 122 receives company X: company X1 and item Y: camera, and then in step S602, reads the BOM information related to the camera. In the example described in the BOM information 133 in Fig. 5, "camera" is searched for in the parent item name column, and its child item name "lens unit" is read.

[0038] FIG. 7 is an explanatory diagram showing an example of a BOM information image based on the BOM information 133 according to this embodiment.

[0039] 7A shows an image of the BOM information for "camera" acquired in step S602. This example shows a camera and a lens unit as a component of the camera. It is conceivable that components other than the one item in this example (i.e., the lens unit) are required for camera production, or that there are cameras that do not use the items in this example. However, since disclosure of these components is not essential for explaining the operation of this embodiment, only the lens unit will be described as an example, and descriptions of other examples will be omitted.

[0040] Fig. 7(b) shows an image of the BOM information for the "lens unit" acquired in step S602 from the second time onwards in the loop process of Fig. 6 (described later). Note that in Figs. 7(a) and 7(b), the left side corresponds to the upstream of the supply chain.

[0041] In step S603, the trading company estimation unit 122 acquires, from the trading history information 132, trading history in which "Company X1" is the ordering company.

[0042] FIG. 8 is an explanatory diagram showing an example of the transaction history information 132 held by the supply chain network creation device 100 according to this embodiment.

[0043] The example of transaction history information 132 shown in Figure 8 includes information such as ID 801, sub-number 802, order receiving company name 803, ordering company name 804, handled item name 805, requested delivery date 806, quantity 807, delivery location 808, delivery date 809, and delivery quantity 810. When the transaction history in which ordering company name 804 is "Company X1" is obtained from the example in Figure 8, the two rows corresponding to ID 801 are "E001" and "E005." In the following explanation, the information in these rows will also be referred to as "transaction history E001" and "transaction history E005."

[0044] The transaction history information 132 is accumulated in business-to-business transactions such as trade and e-commerce services. Depending on the contract for the transaction service, the use of the transaction history information 132 as in this embodiment may be impossible or require permission, but this embodiment is premised on the possibility of use.

[0045] In steps S604 and S605, the trading company inference unit 122 infers from which company the child item acquired in step S602 is being supplied, using the trading history acquired in step S603. Specifically, from the trading history acquired in step S603 where the ordering company name 804 is "Company X1," data where the item name 805 is "Lens Unit" is extracted, and the company registered as the order-receiving company in that data is deemed to be the company supplying the "Lens Unit" to "Company X1." If multiple companies are applicable, it is deemed that multiple companies are supplying the "Lens Unit."

[0046] When comparing items in steps S604 and S605, both item names are converted to standard names and the standard names are compared. The reason for using standard names is that variations in the way item names are written often occur. The item names in the BOM information 133 created by the parts structure generation unit 121 depend on the input information used to train the learning model 131. The item names in the transaction history are defined according to the convenience of the person in charge at the receiving company or the ordering company, and the way item names are written varies even for the same item.

[0047] For example, the actual notation for "lens unit" varies widely, including "Lens U," which is an abbreviation of "unit" with the letter "U," a string of alphanumeric characters such as "XXX-001" indicating the model name of the lens unit, and notations such as "Lens U 112mm," which adds specifications. These notations may show similarities, such as when some of the characters match, as in "lens unit" and "Lens U," but they may also not match at all, as in "lens unit" and "XXX-001."

[0048] Therefore, in this embodiment, these item names are not compared directly, but are converted into standard names individually before comparison. In this embodiment, an example will be described in which product code information is used as the standard name.

[0049] FIG. 9 is an explanatory diagram showing an example of product code information referenced by the trading company inference unit 122 according to this embodiment.

[0050] 9 is a code for classifying items, and a product code 901 and a classification name 902 are registered. In this embodiment, as an example, the product code information will be described using the UNSPSC (United Nations Standard Products and Services Code), a product classification code used internationally. Note that, as long as the item is classified, a standard name defined by a predetermined rule, which is different from the UNSPSC, may be used.

[0051] Returning to the explanation of the processing of each step, in step S604, the trading company inference unit 122 converts the child item name "lens unit" in the BOM information 133 into a standard name, and the item name "camera lens unit assembly" in the transaction history information into a standard name. In both cases, the item name is input into the learning model 131, and the product code is obtained from the output information of the learning model 131. The learning model 131 used in this step is based on the premise that a relationship between information about the item and the standard name (product code) has been established.

[0052] The learning model 131 is generated by performing both learning to generate a BOM in step S202 of Figure 2 and learning to generate a standard name in step S604 of Figure 6 (more specifically, step S6042 of Figure 10, which will be described later).

[0053] In the former learning, for example, when information about an item is input, the system is trained to output the items of items that make up the item (for example, parts that make up a product).

[0054] In the latter case, for example, when a string of characters for an item is input, the robot is trained to output at least one of the standard name of the item, shape information of the item (e.g., a drawing or photograph), and product name of the item (which may include the product category name), and further when at least one of the shape information of the item and the product name of the item is input, the robot is trained to output the standard name of the item.

[0055] Furthermore, in the latter learning, the device may be trained to output at least one of the standard name, shape information, and product name in response to the input of a combination of an item string and at least one of the items other than the item in the transaction history information 132 (for example, the order receiving company name 803 or the ordering company name 804).

[0056] In addition, the learning model 131 may include multiple learning models that have undergone the above-mentioned learning, for example, a first learning model generated by the former learning and a second learning model generated by the latter learning.

[0057] FIG. 10 is a flowchart showing an example of the standard name conversion process performed by the trading company inference unit 122 according to this embodiment.

[0058] 10 illustrates the details of the standard name conversion process in step S604 of FIG. 6. In step S6041, the trading company inference unit 122 creates character strings to be input to the learning model 131. Each character string includes an item name, and the trading company inference unit 122 creates at least one character string. For example, if it is desired to convert the product code from "lens unit" to UNSPSC as the standard name, a character string such as "Please tell me the UNSPSC for lens unit" is created.

[0059] In step S6042, the trading company inference unit 122 inputs the created character string into the learning model 131 and extracts the standard name from the character string output. In the case of "lens unit," a character string including "product code: 45121603 camera lens" is output.

[0060] In step S6043, the trading company inference unit 122 checks whether a standard name has been obtained. As mentioned above, there are no specific rules for writing product names, so the learning model 131 may not always be able to identify a standard name. If a standard name has been identified (step S6043: Yes), the process ends.

[0061] If the standard name cannot be identified (step S6043: No), the trading company inference unit 122 changes the input information to the learning model in step S6044. Here, conversion via shape information rather than character strings and adding information other than the item name make it easier for the learning model to narrow down the characteristics of the item name. Specifically, there are at least two methods: (1) using the learning model 131 to convert the item name into shape information such as an image or drawing, and passing the shape information to step S6042; and (2) generating a character string consisting of a combination of at least one of the item name and item information or transaction history information, and passing the character string to S6042.

[0062] In method (1), the trading company inference unit 122 inputs "Please output the shape of the lens unit" to the learning model 131, acquires an image of the lens unit, and passes the image and the character string "Please tell me the product name and UNSPSC of this image" to step S6042. In method (2), the trading company inference unit 122 passes "Please tell me the product name and UNSPSC of the product M03XLD handled by company X4" to step S6042.

[0063] FIG. 11 is an explanatory diagram showing an example of a process in which the trading company inference unit 122 according to this embodiment converts an input product name into a standard name.

[0064] 11 shows an example of the conversion process from the product name to the output of the standard name in step S604. (1) shows an example of converting "Lens U" to an image and then converting the image to the standard name. (2) shows an example of adding the company name "Company X4" to the model name "M03XLD" to obtain the product name "High Precision Lens," and then converting the product name "High Precision Lens" to the standard name.

[0065] For example, as shown in Figure 11, if the standard name (product code in this example) of a part handled by a certain company under the item name "Lens U" is "45121603," it is desirable to input "Lens U" into learning model 131 and output "45121603." However, if the data learned to create learning model 131 contains sufficient information associating "Lens U" with "45121603," it is considered easy to obtain such an output. However, if the learning data contains insufficient information associating "Lens U" with "45121603," it is expected that such an output will not be obtained.

[0066] Here, for example, if the data learned to create learning model 131 includes information that associates "lens U" with the appearance of that part (for example, a photograph or drawing), and also includes information that associates the appearance of that part with "45121603," then it is thought that by inputting "lens U" into learning model 131 and having the appearance of that part output, and then inputting that appearance into learning model 131, it will be easier to obtain "45121603."

[0067] Similarly, even if the part model name "M03XLD" is input into learning model 131 and the output "45121603" is not obtained, if the learning data contains information that associates the model name "M03XLD" with the part's product name (which may include the product category name) "High Precision Lens," and also contains information that associates "High Precision Lens" with "45121603," then "M03XLD" can be input into learning model 131 to output the appearance of the part, and then that appearance can be input into learning model 131 to obtain "45121603."

[0068] Furthermore, if the learning data contains information that associates a combination of multiple items included in the transaction history information 132, such as an item name and a company name, with the standard name, the appearance of the part, or the product name, etc., inputting that combination into the learning model is expected to improve the accuracy of estimation. For example, when an abstract model name such as "M03XLD" is input, there is a possibility that various companies have assigned similar model names to completely different parts. Therefore, adding the name of the company that received the order (i.e., the name of the company that supplies the part), such as "Company X4," as input to the learning model 131 is expected to make it easier to obtain the desired output.

[0069] The processing from step S605 onwards will be described with reference to Fig. 6 again. Here, as an example, a case will be described in which "lens unit" is specified in the BOM information in step S601, transaction history E001 and transaction history E005 in which the ordering company name 804 is "Company X1" are acquired in step S603, and the following standard name is acquired as a processing result in step S604.

[0070] ・The standard name of the "lens unit" in the BOM information is "45121603". ・The standard name of the "camera lens unit" in transaction history E001 is "45121603". ・The standard name of the "kenbikyo" in transaction history E005 is "41111710".

[0071] In step S605, the trading company inference unit 122 searches for the item's parts from the trading history using the standard name. The trading company inference unit 122 compares the standard name "45121603" of "lens unit" with the two rows of IDs E001 and E005 extracted in step S603, and obtains the order-receiving company "Company X2" with ID E001 that matches the standard name "45121603" of "camera lens unit assembly."

[0072] In step S606, the trading company inference unit 122 outputs the extracted trading history to the SC (supply chain) network information 134.

[0073] FIG. 12 is an explanatory diagram showing an example of the SC network information 134 held by the supply chain network creation device 100 according to this embodiment.

[0074] As shown in Figure 12, the SC network information stores information about an ordering company name 1201, an order-receiving company name 1202, a transaction item name 1203, and a transaction product code 1204. The information about the "lens unit" is registered as shown in the first line of Figure 12, with the ordering company name 1201: Company X1, the order-receiving company name 1202: Company X2, the transaction item name 1203: "camera lens unit assembly," and the transaction product code 1204: 45121603. If no corresponding transaction history is found in step S605, it is assumed that the ordering company is producing the item, and the ordering company is registered in the order-receiving company name 1202.

[0075] In step S607, the trading company inference unit 122 passes the order receiving company: Company X2 and the child item: "Lens Unit" to step S601 as targets for creating the next supply chain network. Note that if there are no targets for creating the next supply chain network, the process ends.

[0076] The trading company inference unit 122 returns to step S601, and in step S602 performs the processes of steps S603 to S606 for each of the three items, "lens," "AF motor (autofocus motor)," and "aperture," based on the BOM information for "lens unit" (illustrated in FIG. 7B). As described above, this process is performed for all items linked from the BOM information for "camera."

[0077] FIG. 13 is an explanatory diagram showing an example of a supply chain network display screen output by the supply chain network creation device 100 according to this embodiment.

[0078] The display screen S1300 is composed of a BOM estimation result S1301, a supply chain network estimation result S1302, and estimation details S1303. The display screen S1300 is displayed on the display unit 160 when a company and an item name to be displayed in the supply chain network are specified from the input unit 110. Fig. 13 shows an example of the display when the company: Company A, and the item: item 1 are entered.

[0079] The BOM estimation result S1301 displays, in a tree format, BOM information 133 of items linked from item1 created by the parts structure generation unit 121. The supply chain network estimation result S1302 displays, in a tree format, SC network information 135 created by the trading company estimation unit 122.

[0080] When a user operating this screen selects a circular image in the tree via the input unit 110, detailed information is displayed in the estimation details S1303. FIG. 13 shows an example of the display when item 3, an image of company B (displayed as a black circle in the BOM estimation result S1301 and the supply chain network estimation result S1302), is selected. The user can check whether the estimation details are correct and select either the correct or incorrect check box in the user evaluation field 1314. The selection result is stored in the memory unit 130 and, if necessary, is input as learning information for the learning model.

[0081] In the example of Figure 13, by inputting the item name "item3" included in the estimation results of the BOM information into the learning model 131, the appearance of the part of that item (for example, the shape identified by a photograph or drawing) is estimated, and by inputting this appearance into the learning model 131, the item name "item3" is estimated. In contrast, by inputting the item name "item-003" included in Company B's transaction history into the learning model 131, the appearance of the part of that item is estimated, and by inputting this appearance into the learning model 131, the item name "item3" is estimated. From these estimation results, it is estimated that both are parts of the same item.

[0082] 13 , when the estimation result is presented to the user and the user enters an evaluation result (i.e., a judgment result on the accuracy of the estimation result) in a user evaluation field 1314, the estimation result and information indicating its accuracy are added to the learning data for generating the learning model 131. This process may be performed, for example, when step S6043: Yes.

[0083] In the above example, the user evaluates the estimation result of the standard name, but the user's evaluation results may also be obtained for other estimation results (e.g., estimation results of the BOM) by the learning model 131 in a similar manner, and these results may be added to the learning data. By learning the updated learning data in this way, the accuracy of estimation by the learning model 131 is improved.

[0084] [Hardware Configuration] FIG. 14 is a block diagram showing an example of the hardware configuration of the supply chain network creation device 100 according to this embodiment.

[0085] In this embodiment, the supply chain network creation device 100 is, for example, an information processing device such as a personal computer, a workstation, or a server device, but is not limited to these and may also be an electronic information terminal such as a smartphone, a mobile phone terminal, or a PDA (Personal Digital Assistant).

[0086] In other words, the supply chain network creation device 100 has an input device 201, an output device 202, an external memory device 203, an arithmetic device 204, a main memory device 205, a communication device 206, and a bus 207 that connects each device to each other.

[0087] The input device 201 is a device that accepts input from, for example, a keyboard, a mouse, a touch pen, or other pointing devices. The output device 202 is a device that displays information, for example, a display. The external storage device 203 is a non-volatile storage device, for example, a hard disk drive or flash memory. The arithmetic device 204 is a arithmetic device, for example, a CPU (Central Processing Unit).

[0088] The main memory device 205 is a memory device such as a RAM (Random Access Memory). The communication device 206 is a wireless communication device that performs wireless communication via an antenna, or a wired communication device that performs wired communication via a network cable. The memory unit 130 of the supply chain network creation device 100 is realized by the main memory device 205 or the external memory device 203 of the supply chain network creation device 100.

[0089] In addition, the input unit 110, calculation unit 120, communication unit 140, output unit 150 and display unit 160 of the supply chain network creation device 100 are realized by a program that causes the calculation unit 204 of the supply chain network creation device 100 to perform processing.

[0090] This program is stored in the main memory device 205 or the external memory device 203, and is loaded onto the main memory device 205 for execution, and executed by the arithmetic unit 204. Furthermore, the supply chain network creation device 100 may have a communication device 206 as needed, and may send and receive information via an external network connected to the communication device 206.

[0091] Furthermore, the system according to the embodiment of the present invention may be configured as follows.

[0092] (1) A supply chain network creation method using a computer system, wherein the computer system (e.g., supply chain network creation device 100) has an arithmetic unit (e.g., arithmetic unit 204) and a storage device (e.g., main storage device 205 and external storage device 203), the storage device holds a learning model (e.g., learning model 131) and transaction history information (e.g., transaction history information 132), the learning model is trained to output information indicating item numbers of items constituting the target item in response to input of information about the target item, the transaction history information including information associating the item numbers with a company that has ordered the item and a company that has received an order for the item, and the supply chain network creation method is performed by the arithmetic unit The method includes a first step (e.g., processing by the parts structure generation unit 121, FIG. 2) of using the learning model to estimate the items of items that make up the item based on information about the input target item; a second step (e.g., processing by the trading company estimation unit 122, step S604 in FIG. 6 and processing by the trading company estimation unit 122, step S605 in FIG. 6) of converting a character string of the estimated item and a character string of an item included in the transaction history information into names (e.g., standard names) defined by predetermined rules by the calculation device; and a third step (e.g., processing by the trading company estimation unit 122, step S605 in FIG. 6) of comparing the estimated item with an item included in the transaction history information based on the converted names by the calculation device to estimate a supplier company of the estimated item.

[0093] This makes it possible to create a supply chain network related to the production of a product, all the way up to upstream suppliers.

[0094] (2) In the supply chain network creation method described in (1) above, the learning model is further trained to output at least one of a name defined by the predetermined rule, shape information of the item, and a product name of the item when a character string of the item is input, and to output a name defined by the predetermined rule when at least one of the shape information of the item and the product name of the item is input; in the second step, the computing device inputs the character string of the item into the learning model to obtain the name defined by the predetermined rule (e.g., steps S6042 and S6043: Yes); if the name defined by the predetermined rule cannot be obtained but at least one of the shape information of the item and the product name of the item can be obtained (e.g., step S6043: No, step S6042 after changing the input information in step S6044), the computing device inputs at least one of the shape information of the item and the product name of the item into the learning model to obtain the name defined by the predetermined rule.

[0095] This allows for the accuracy of supplier estimation to be improved by obtaining the standard name through conversion to other information even when the standard name cannot be obtained directly.

[0096] (3) The supply chain network creation method described in (2) above, wherein the learning model is further trained to output at least one of a name defined by the predetermined rule, shape information of the item, and product name of the item in response to an input of a combination of a character string of the item and at least one item other than the item in the transaction history information, and in the second step, the computing device inputs a combination of a character string of the item and at least one item other than the item (e.g., a company name) in the learning model to obtain at least one of a name defined by the predetermined rule, shape information of the item, and product name of the item.

[0097] This allows for more accurate conversion to standard names.

[0098] (4) In the supply chain network creation method described in (1) above, the calculation device starts the first step when it receives input of the item name of the target item as information about the target item.

[0099] This allows the creation of a supply chain network for the desired item.

[0100] (5) The supply chain network creation method described in (1) above further includes a fourth step in which the computing device outputs an estimation result of the item and the supplier company of the item, and when information indicating whether the estimation result is correct or incorrect is input, the estimation result of the item and the supplier company of the item and the information indicating whether the estimation result is correct or incorrect are added to the data learned by the learning model.

[0101] This improves the accuracy of estimation based on the learning model.

[0102] The present invention is not limited to the above-described embodiments and includes various modifications. For example, the above-described embodiments have been described in detail to provide a better understanding of the present invention, and are not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

[0103] Furthermore, the above-described configurations, functions, processing units, processing means, etc. may be partially or entirely implemented in hardware, for example, by designing them as integrated circuits. The above-described configurations, functions, etc. may also be implemented in software, with a processor interpreting and executing a program that implements each function. Information such as programs, tables, and files that implement each function can be stored in storage devices such as nonvolatile semiconductor memory, hard disk drives, and solid-state drives (SSDs), or in computer-readable, non-transitory data storage media such as IC cards, SD cards, and DVDs.

[0104] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. In reality, it can be considered that almost all components are interconnected.

[0105] Furthermore, the technical elements of the above-described embodiments may be applied independently, or may be applied by dividing them into multiple parts such as program parts and hardware parts.

Claims

1. A supply chain network creation method using a computer system, the computer system having an arithmetic unit and a storage device, the storage device holding a learning model and transaction history information, the learning model having been trained to output information indicating item names of items that make up a target item in response to input of information about the target item, the transaction history information including information correlating item names with a company that has ordered the item and a company that has received an order for the item, the supply chain network creation method comprising the steps of: a first step in which the arithmetic unit uses the learning model to estimate item names of items that make up the target item based on the input information about the target item; a second step in which the arithmetic unit converts character strings of the estimated item and character strings of items included in the transaction history information into names defined in accordance with predetermined rules; and a third step in which the arithmetic unit compares the estimated item with items included in the transaction history information based on the converted names to estimate a supplier company of the estimated item.

2. A supply chain network creation method as described in claim 1, wherein the learning model is further trained to output at least one of a name defined by the predetermined rule, shape information of the item, and product name of the item when a character string of the item is input, and to output a name defined by the predetermined rule when at least one of the shape information of the item and product name of the item is input, and wherein in the second step, the computing device obtains the name defined by the predetermined rule by inputting the character string of the item into the learning model, and if the name defined by the predetermined rule cannot be obtained but at least one of the shape information of the item and product name of the item can be obtained, the computing device obtains the name defined by the predetermined rule by inputting at least one of the shape information of the item and product name of the item into the learning model.

3. A supply chain network creation method as described in claim 2, wherein the learning model is further trained to output at least one of a name defined in the predetermined rule, shape information of the item, and product name of the item in response to an input of a combination of a character string of the item and at least one item other than the item in the transaction history information, and wherein in the second step, the computing device inputs a combination of a character string of the item and at least one item other than the item in the transaction history information into the learning model, thereby obtaining at least one of a name defined in the predetermined rule, shape information of the item, and product name of the item.

4. A supply chain network creation method as described in claim 1, characterized in that the calculation device starts the first step when it receives input of the item name of the target item as information related to the target item.

5. A supply chain network creation method as described in claim 1, further comprising a fourth step in which the computing device outputs an estimation result of the item and the supplier company of the item, and when information indicating whether the estimation result is correct or incorrect is input, adds the estimation result of the item and the supplier company of the item and the information indicating whether the estimation result is correct or incorrect to the data learned by the learning model.

6. A supply chain network creation device comprising: an arithmetic unit; and a storage device; wherein the storage device holds a learning model and transaction history information; wherein the learning model has been trained to output information indicating the item names of items that make up a target item in response to input of information about the target item; wherein the transaction history information includes information that associates the item names with a company that has ordered the item and a company that has received an order for the item; wherein the arithmetic unit uses the learning model to estimate the item names of items that make up the target item based on the input information about the target item; converts character strings of the estimated item and character strings of items included in the transaction history information into names defined by predetermined rules; and compares the estimated item with items included in the transaction history information based on the converted names, thereby estimating a supplier company of the estimated item.

7. A supply chain network creation device as described in claim 6, wherein the learning model is further trained to output at least one of a name defined in the predetermined rule, shape information of the item, and product name of the item when a character string of the item is input, and to output a name defined in the predetermined rule when at least one of the shape information of the item and product name of the item is input, and wherein the calculation device obtains the name defined in the predetermined rule by inputting the character string of the item into the learning model, and when the name defined in the predetermined rule cannot be obtained but at least one of the shape information of the item and product name of the item can be obtained, inputs at least one of the shape information of the item and product name of the item into the learning model to obtain the name defined in the predetermined rule.

8. A supply chain network creation device as described in claim 7, wherein the learning model is further trained to output at least one of a name defined in the predetermined rule, shape information of the item, and product name of the item in response to an input of a combination of a character string of the item and at least one item other than the item in the transaction history information, and the calculation device obtains at least one of a name defined in the predetermined rule, shape information of the item, and product name of the item by inputting a combination of a character string of the item and at least one item other than the item in the transaction history information into the learning model.

9. A supply chain network creation device as described in claim 6, characterized in that when the calculation device receives input of the item of the target item as information about the target item, it begins to estimate the items of the items that make up the item using the learning model.

10. A supply chain network creation device as described in claim 6, wherein the calculation device outputs an estimation result of the item and the supplier company of the item, and when information indicating whether the estimation result is correct or incorrect is input, the estimation result of the item and the supplier company of the item and the information indicating whether the estimation result is correct or incorrect is added to the data learned by the learning model.

11. A program for causing a computer system to create a supply chain network, the computer system comprising an arithmetic unit and a storage device, the storage device holding a learning model and transaction history information, the learning model having been trained to output information indicating the item names of items that make up a target item in response to input of information about the target item, the transaction history information including information correlating the item names with the company that has ordered the item and the company that has received an order for the item, the program causing the arithmetic unit to execute the following steps: a first step of using the learning model to estimate the item names of items that make up the target item based on the input information about the target item; a second step of converting character strings of the estimated item and character strings of items included in the transaction history information into names defined by predetermined rules; and a third step of estimating a supplier company of the estimated item by comparing the estimated item with items included in the transaction history information based on the converted names.

Images