Order placement support device, order placement support method and order placement support program

Abstract

To provide an order placement support device and the like which can: be used for purchasing a facility or a component; reduce time and labor until order placement of an order placement object is completed; and efficiently find an order placement destination from which the order placement object can be purchased at the lowest price.SOLUTION: An order placement support device is configured to: on the basis of analysis information acquired by performing latent semantic analysis on a purchase specification α2 for specifying an order placement object, automatically apply object kind information of a new order placement object; extract a customer candidate which can be selected on the basis of the object kind information; and present purchase conditions matching the object kind. The order placement support device is further configured to: estimate, for each customer candidate, a reduction ratio indicating a ratio at which a price is reduced to a contract amount from a budget amount; and display, as a final customer candidate, the customer candidate having a high rank of the estimated reduction ratio. Therein the order placement support device estimates the reduction rate by using a learning model 10 obtained by causing the model to machine-learn a correlation between input data which includes, the object kind information, customer information, and the purchase conditions, and output data including the reduction ratio, on the basis of the object kind information of the new order placement object; the presented purchase conditions or purchase conditions which are finally selected by referring to the presented purchase conditions, and customer information of each extracted customer candidate.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to an ordering support device, an ordering support method, and an ordering support program that can be used when ordering equipment, parts, etc. from a supplier. [Background technology]

[0002] 2. Description of the Related Art Conventionally, a support device for purchasing equipment, parts, and the like from a supplier is known, as disclosed in Patent Document 1.

[0003] This comprises an acquisition unit that acquires multiple pieces of order data including first information indicating the budget for the order target, second information indicating the actual cost of the order target, third information indicating the time when the order target was ordered, fourth information indicating the delivery date of the order target, and amount information indicating the relationship between the budget for the order target and the actual cost of the order target, and a model generation unit that generates a tree-structured learning model using the first to fourth pieces of order data as independent variables and the amount information as a dependent variable, thereby presenting order conditions for new order targets that take past purchasing history into consideration, thereby reducing ordering costs. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2022-91337 A Summary of the Invention [Problem to be solved by the invention]

[0005] By using such a system, even beginners with little ordering experience can make purchases in the same way as experienced veterans, which is expected to effectively reduce procurement costs.

[0006] However, when purchasing equipment or parts, a procurement request is made by creating a purchase specification and sending it to the procurement department, but it is not easy to understand the qualitative content of the purchase specification, and it takes time for the procurement department to thoroughly examine the content of the purchase specification and identify what the order is for. In addition, when deciding on the supplier that can purchase the product at the lowest price, a certain amount of advance research and preparation is required, and even veterans in the procurement department need a certain amount of time and effort to place an order for the product.

[0007] The present invention has been made in consideration of the above circumstances, and has as its main objective the provision of an ordering support device, ordering support method, and ordering support program that can be used by anyone, from beginners to veterans, when procuring equipment, parts, etc., that can reduce the time and effort required to order an item, and that can accurately find a supplier that is likely to be able to purchase items at the lowest price. [Means for solving the problem]

[0008] In order to achieve the above object, an ordering support device according to the present invention comprises: An ordering support device that supports ordering of an order target, an analysis information acquisition unit that performs latent semantic analysis on a purchase specification created to identify an order target and acquires analysis information of the purchase specification; a product type information database that stores analysis information acquired by the analysis information acquisition unit for past purchase specifications and product type information assigned to an order target identified in the past purchase specifications in association with each other; A business partner information database for storing business partner information; a purchase condition presentation unit that presents purchase conditions that match the product type; a learning model storage unit that stores a learning model that has been machine-learned to learn a correlation between input data including the product type information, the customer information, and the purchasing conditions, and output data including a reduction rate that indicates a reduction rate from the budget amount to the contract amount; Equipped with A product type information assignment unit that acquires product type information of the new order target by referring to the product type information database based on analysis information obtained by performing latent semantic analysis on the purchase specification of the new order target by the analysis information acquisition unit; a supplier candidate extraction means for extracting selectable supplier candidates by referring to the supplier information database based on the product type information assigned by the product type information assignment unit; a reduction rate estimation means for estimating a reduction rate for each of the supplier candidates using the learning model from the product type information assigned by the product type information assignment unit, the purchasing conditions presented by the purchasing condition presentation unit for the product type to be newly ordered or final selected purchasing conditions finally selected with reference thereto, and supplier information of each of the supplier candidates extracted by the supplier candidate extraction means; a final supplier candidate display means for displaying the supplier candidates with the top reduction rates estimated by the reduction rate estimation means as final supplier candidates; The present invention is characterized in that it has the following features.

[0009] Here, the display of the final trading partner candidates may be such that each of the trading partners with the top reduction rates is displayed together with the corresponding reduction rate, or such that the trading partners with the top reduction rates are displayed together with the largest reduction rate (expected reduction rate).

[0010] Therefore, after obtaining a purchase specification that specifies the order target, the ordering person performs a latent semantic analysis of the purchase specification using the analysis information acquisition unit, and acquires analysis information (LSA similarity) of the purchase specification.

[0011] Then, the product type information assigning unit assigns product type information of the order target by referring to the product type information database based on the analysis information (LSA similarity), The supplier candidate extraction means extracts selectable supplier candidates by referring to the supplier information database based on the product type information assigned by the product type information assignment unit.

[0012] At this stage, the primary supplier candidates are extracted only from the product type information, so suppliers with little or no expected reduction rate will be extracted.

[0013] Thereafter, the reduction rate estimation means estimates a reduction rate for each of the candidate suppliers using a learning model from the product type information assigned by the product type information assignment unit, the purchasing conditions presented by the purchasing condition presentation unit or the final selected purchasing conditions finally selected with reference thereto, and the supplier information of each of the candidate suppliers extracted by the candidate supplier extraction means.

[0014] Then, the final supplier candidate display means displays the supplier candidates with the top reduction rates estimated by the reduction rate estimation means as final supplier candidates.

[0015] Therefore, since the ordering staff member can automatically obtain information about the type of product being ordered from the purchasing specifications, he or she no longer needs to analyze the specifications himself or herself to identify the item being ordered (the type of product being purchased).

[0016] In addition, potential suppliers are extracted from product type information, and purchasing conditions that match the product type are presented. Using a learning model, the reduction rate of each potential supplier is estimated from the product type information, the presented purchasing conditions or the final selected purchasing conditions that are finally selected based on the product type information, and the supplier information of each of the extracted potential suppliers. The potential supplier with the highest reduction rate is displayed as the final potential supplier, so that the purchaser can look at the content and place an order with the most suitable potential supplier (for example, the one with the highest reduction rate), or have multiple potential suppliers with the highest reduction rates submit competitive bids. This allows appropriate ordering procedures to be carried out regardless of business experience, and also makes it possible to reduce ordering costs.

[0017] Here, the purchasing condition presentation unit may create a tree-structured learning model for each product type, in which the purchasing conditions and the supplier information are independent variables and the reduction rate is a dependent variable, identify a decision tree linked to the product type information assigned by the product type information assignment unit, and present the purchasing conditions of the decision tree.

[0018] In other words, the purchasing condition presentation unit may use a tree-structured learning model in which the purchasing conditions and supplier information are independent variables and the reduction rate is the dependent variable, thereby presenting purchasing conditions that are expected to reduce costs when ordering the item to be ordered. Effect of the Invention

[0019] As described above, based on the analysis information obtained by performing latent semantic analysis on the purchasing specifications, the product type information of the order target is automatically assigned by referring to the product type information database in which the analysis information and product type information are stored in association with each other, thereby reducing the time it takes for a procurement department staff member to identify the order target.

[0020] In addition, candidate suppliers are extracted from product type information, and purchasing conditions that match the product type are presented. The reduction rate of each candidate supplier is predicted using a learning model that predicts the reduction rate from the product type information, the presented purchasing conditions or the final selected purchasing conditions finally selected based on this, and the supplier information of each extracted candidate supplier, and the candidate suppliers with the highest reduction rates are displayed as the final candidate suppliers, so that ordering personnel, regardless of experience (from beginners to veterans), can effectively find suppliers who can purchase at low cost and place orders. [Brief description of the drawings]

[0021] [Figure 1] 1 is a schematic configuration diagram showing an order support device according to the present invention; [Diagram 2] FIG. 11 is an explanatory diagram illustrating an example of assigning product type labels using the LSA technique. [Diagram 3] FIG. 1 is a block diagram showing an example of a machine learning device of an ordering support device according to the present invention. [Figure 4] FIG. 1 is a diagram showing an example of learning data (input data and output data) used in machine learning. [Diagram 5] 1 is a flowchart showing an example of a machine learning method performed by the machine learning device according to the present invention. [Figure 6] 1 is a block diagram showing an example of a supplier proposal device of an ordering support device according to the present invention; [Figure 7] FIG. 1 is a diagram showing an example of a supplier information database in which business items (items handled) and purchasing records of suppliers are recorded. [Figure 8] 13 is a flowchart showing an example of a supplier suggestion method by the supplier suggestion device. [Figure 9] FIG. 13 is a diagram showing a display example of a final supplier candidate display means, showing an example in which the contents presented by the ordering support device and the contents finally selected by the ordering person are displayed. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0023] FIG. 1 shows a schematic configuration of an order support device 1 according to the present invention.

[0024] As shown in the figure, the ordering support device 1 includes an analysis information acquisition unit (an analysis information acquisition unit 2a used in the learning phase and an analysis information acquisition unit 2b used in the inference phase), a product type information assignment unit 3 used in the inference phase, a product type information database 4, a machine learning device 5, a purchasing condition presentation unit 6, a supplier candidate extraction unit 8, a supplier information database 7, a supplier proposal device 9, and a display unit 11.

[0025] The analysis information acquisition units 2a and 2b acquire analysis information (LSA similarity) of the purchase specifications (purchase specification α1 used in the learning phase, purchase specification α2 used in the inference phase) by performing a well-known analysis using a technique called LSA (Latent Semantic Analysis).

[0026] The analysis information (LSA similarity) obtained here is calculated by the similarity between the vector obtained by latent semantic analysis of the purchasing specification and the vector obtained by latent semantic analysis of other purchasing specifications, and for example, the cosine value (cosine similarity) between the purchasing specifications is used. In other words, an index is used of the extent to which the vectors of the purchasing specifications are oriented in the same direction.

[0027] Purchase specifications α1 and α2 are documents that describe the requirements for ordering the equipment, devices, parts, etc. that are the subject of the order, and contain various information about the order target, such as the name and quantity of the purchased item, equipment specifications, auxiliary equipment, installation location, etc. Purchase specifications α1 are purchase specifications that were used when placing a past order, and purchase specifications α2 are purchase specifications that were created to specify a new order target.

[0028] It takes time for a procurement department staff member to carefully examine such purchase specifications and determine what the order target is from the purchase specifications, and if the staff member is not accustomed to this, there is a risk of making a mistake in the order target. Therefore, the analysis information acquisition units 2a and 2b perform latent semantic analysis on the purchase specifications α1 and α2, and perform natural language processing using a vector space model to identify the order targets described in the purchase specifications α1 and α2 in the form of LSA similarity.

[0029] In this case, the entire purchase specifications α1 and α2 may be subjected to latent semantic analysis, or only the part of the purchase specifications in which the contents of the order target are directly described may be subjected to latent semantic analysis.

[0030] The product type information is information for identifying the order target based on the vector of the purchase specification acquired by the analysis information acquisition unit 2a, and is composed of product type labels manually assigned in the learning phase. This manually assigned product type information constitutes a part of the input data as learning data for forming a learning model described later.

[0031] The product type information used in the learning phase is given as follows.

[0032] As shown in Figure 2, based on the LSA similarity (the cosine value between the vector of the purchasing specification obtained by LSA analysis and the vector obtained by latent semantic analysis of other purchasing specifications) acquired by the analysis information acquisition unit 2a, similar purchasing specifications (those with cosine values ​​within a certain range) are extracted, and these purchasing items are grouped together and provisionally assigned a product type label that represents a higher conceptualization.

[0033] For example, when looking at a purchasing specification (purchased item) with an LSA similarity of 0.90 or more with a certain specification as a result of analyzing multiple past specifications, if the purchased item in one purchasing specification is a slurry transport equipment and the purchased item in another purchasing specification is a fly ash silo, these are combined and provisionally assigned the product type label of the superordinate concept "ash handling equipment."Also, when looking at specifications with other similar vector groups, if the purchased item in one purchasing specification is a "strainer" and the purchased item in the other purchasing specification is a "sand separator," the product type information combining these is provisionally assigned the product type label of the superordinate concept "filtration equipment."

[0034] After that, the angle between the vectors of each purchasing specification that has been provisionally assigned the same product type label is calculated, and it is confirmed that they are similar to each other (if there are any purchasing specifications that have low similarity, the label is removed).

[0035] This process is repeated until there are no more purchasing specifications that do not have product type labels attached, and then the provisional product type labels are confirmed as official labels.

[0036] Then, the LSA similarity acquired by the analysis information acquisition unit 2a and the product type information assigned thereto are associated with each other and stored in the product type information database 4. That is, the product type information database 4 registers, for each of a large number of past purchase specifications 10a, the LSA similarity acquired by LSA (latent semantic analysis) and the product type information assigned to each LSA similarity in a linked manner.

[0037] In addition, the product type may be clarified by adding the representative product name and construction name included in the corresponding vector group as supplementary information for the product type information (product type label). For example, the information "item names included in the ash handling equipment: slurry transport equipment, fly ash silo" is added as supplementary information for the above "ash handling equipment", and the information "item names included in the filtration equipment: strainer, sand separator" is added as supplementary information for the above "filtration equipment".

[0038] By using the product type information database 4 that stores the LSA similarity of the order target in association with the product type information, it is possible to obtain the product type information for the LSA similarity by acquiring the LSA similarity for the purchase specification of the order target to be newly ordered in the future. Therefore, the product type information assigning unit 3 used in the inference phase automatically assigns the corresponding product type information by referring to the product type information database 4 based on the LSA similarity acquired by the analysis information acquiring unit 2b.

[0039] The machine learning device 5 operates as the subject of the learning phase, and generates a learning model 10 used to infer the reduction rate of the order target by machine learning. That is, the machine learning device 5 uses as input data the product type information (product type label) assigned to the analysis information (LSA similarity) obtained by performing latent semantic analysis on the purchase specification α1 of the order target that was previously ordered, the purchase conditions at the time of purchase that can be understood from the purchase specification, and the supplier information, and obtains many sets of learning data in which the reduction rate indicating the reduction rate from the budget amount to the contract amount ((budget amount-contract amount) / budget amount) is output as output data, and performs machine learning on the correlation between these input data and output data to form the learning model 10 (this point will be described in detail).

[0040] The purchasing condition presenting unit 6 presents reference purchasing conditions for final selection of purchasing conditions that are part of the input data when estimating the reduction rate using the learning model.

[0041] The supplier information database 7 is a database in which various information related to suppliers is stored, and the supplier candidate extraction unit 8 refers to the supplier information database 7 and extracts supplier candidates based on the product type information assigned by the product type information assignment unit 3.

[0042] The supplier proposal device 9 operates as the subject of the inference phase, and uses the trained learning model 10 generated by the machine learning device 5 to infer the reduction rate of the order target from the product type information assigned by the product type information assignment unit 3, the final selected purchasing conditions finally selected based on the purchasing conditions presented by the purchasing condition presentation unit 6, and the supplier information of candidate suppliers extracted by referring to the supplier information database 7, and proposes suppliers with the highest reduction rates.

[0043] The display unit 11 can be substituted by the display screen of various terminals, and displays information on the final business partner candidates proposed by the business partner proposal device 9, information on the reduction rate, and the like.

[0044] The machine learning device 5 is configured with a general-purpose or dedicated computer. This computer may be a stationary computer or a portable computer, and may also be a client computer, a server computer, or a cloud computer.

[0045] As shown in FIG. 3, the machine learning device 5 includes a learning data acquisition unit 51, a learning data storage unit 52, a machine learning unit 53, and a trained model storage unit 54.

[0046] The learning data acquisition unit 51 functions as an interface unit that acquires learning data including input data and output data via the communication network 12. This learning data acquisition unit 51 is connected via the communication network 12 to the analysis information acquisition unit 2a, the product type information database 4, the worker terminal 13 used by the worker, and the like.

[0047] The learning data storage unit 52 is a database that stores a plurality of sets of learning data acquired by the learning data acquisition unit 51 .

[0048] The machine learning unit 53 performs machine learning using the learning data stored in the learning data storage unit 52. The machine learning unit 53 inputs a plurality of sets of learning data to the learning model 10, thereby causing the learning model 10 to learn the correlation between the input data included in the learning data and the output data corresponding to the input data, thereby generating a learned model.

[0049] There are various methods for generating a trained model, but it is preferable to use a neural network, for example, which performs supervised learning by the machine learning unit 53.

[0050] The trained model storage unit 54 is a database that stores the trained learning model 10 generated by the machine learning unit 53. The learning model 10 stored in the trained model storage unit 54 is provided to a supplier proposal device 9 (described later) via an arbitrary communication network or storage medium.

[0051] The learning data storage unit 52 and the trained model storage unit 54 may be configured as a single storage device or as separate storage devices.

[0052] As shown in Fig. 4, the learning data includes, as input data, at least product type information (product type label) manually assigned based on analysis information (LSA similarity) of the purchase specification obtained by performing latent semantic analysis on the purchase specification α1 used at the time of past orders, the purchasing conditions at the time of past orders, and the supplier information with which the transaction was made at that time. In addition, as output data (teaching data), it includes at least a reduction rate indicating the rate at which the budget amount was reduced to the contract amount ((budget amount - contract amount) / budget amount).

[0053] Here, the purchase conditions are conditions at the time of purchase (order) that affect the reduction rate, and include the following information. (1) Competitive or special assignment The reduction rate also varies depending on whether the order is a special order or a competitive bid, and generally, the reduction rate tends to increase with competitive bids. (2) Contract method Even when conducting competitive bidding, it is possible to increase the reduction rate by improving the process. (3) Budget price The size of the budget affects the willingness of the contracting party, which in turn affects the reduction rate. (4) Contract month The reduction rate tends to increase when the contract month is the fiscal year end. (5) Payment terms The reduction rate may be increased by offering more favorable payment terms to business partners. (6) The difference between the quotation request date and the delivery date If there is a margin of error in the delivery schedule, the reduction rate tends to increase. (7) Delivery destination (○○ power plant, ○○ substation, etc.) The reduction rate is affected by local conditions in the delivery area. (8) Contracting Officer The reduction rate is affected by the skill of the individual person in charge. (9) Technical supervisor Differences in budget planners affect reduction rates (10) Age of the contracting officer The more experienced you are, the higher the reduction rate is expected to be. These purchasing conditions are closely related to the reduction rate, so in order to improve the accuracy of the learning model, it is useful to take these purchasing conditions into account and have the model learn the correlation with the reduction rate.

[0054] The business partner information is information about business partners that affect the reduction rate, and includes the following information: (1) Name of Contracting Party Although the reduction rate varies depending on the business partner, it is possible to identify companies that tend to have high reduction rates and companies that tend to have low reduction rates based on past contracts, so identifying the contract partner's name is useful in estimating the reduction rate. (2) Contractor classification (major manufacturers, small construction companies, etc.) Whether the contracting party is a major company or not also affects the reduction rate. (3) Number of years of business with the contracting party, etc. New companies trying to enter the market tend to have higher reduction rates, and companies with long-standing relationships also tend to have higher reduction rates.

[0055] In this way, since the supplier information has a significant effect on the reduction rate, it is useful to take the supplier information into account and have the correlation with the reduction rate learned.

[0056] (Machine learning methods) The machine learning device 5 forms a learning model using, for example, a neural network model. This neural network model is a publicly known model, and learns the correlation between the input data and the output data by inputting input data included in the learning data to an input layer and comparing output data output from an output layer as an estimation result with output data (teacher data) included in the learning data.

[0057] First, as a preliminary preparation for having the machine learning device 5 perform machine learning, the learning data acquisition unit 51 acquires a desired number of learning data and stores the acquired multiple sets of learning data in the learning data storage unit 52. The number of learning data prepared here is appropriately set in consideration of the estimation accuracy required for the learning model 10.

[0058] The learning data can be prepared in various ways. For each past contract subject, the analysis information acquisition unit 2a performs latent semantic analysis on the purchase specification of the contract subject to acquire analysis information (LSA similarity) of the purchase specification, assigns product type information that specifies the order target specified in the purchase specification to this analysis information (LSA similarity) (enter and assign product type labels via the operator terminal 13), links the analysis information and product type information, and stores them in the product type information database 4. At the same time, purchase conditions (contract date, competitive classification, etc.) and business partner information are acquired from the purchase specification α1, and the reduction rate of the contract is calculated from the actual contract amount, and these are input via the operator terminal 13, and are associated with the contract subject, etc. to form a set of learning data. Similar learning data is then created for each past contract subject, and stored in the learning data storage unit 52.

[0059] After the above-mentioned preparations, the machine learning unit 53 performs, for example, a series of processes in the flowchart shown in FIG. 5 to generate a trained model.

[0060] In step S100, the machine learning unit 53 acquires one set of learning data from the multiple sets of learning data stored in the learning data storage unit 52. The learning data may be acquired in a predetermined order or randomly.

[0061] Next, in step S110, the machine learning unit 53 performs supervised machine learning. That is, the machine learning unit 53 inputs input data included in the acquired set of learning data to the prepared learning model 10, outputs an estimated result, compares this estimated result with output data (teacher data) included in the learning data acquired in step S100, and performs machine learning by a known neural network method. In this way, the machine learning unit 53 causes the learning model 10 to learn the correlation between the input data and the output data (reduction rate).

[0062] Thereafter, using this learning model under learning, the above-mentioned operations are performed on the remaining prepared learning data, and learning is continued in the machine learning unit 53. At this time, the machine learning unit 53 determines whether or not to continue machine learning based on the error between the output data and the teacher data, the number of learnings, the remaining amount of unlearned learning data stored in the learning data storage unit 52, etc. (step S120).

[0063] That is, if the machine learning unit 53 determines in step S120 to continue the machine learning (YES), it performs the processes of steps S100 to S110 on the learning model 10 being trained using untrained learning data. On the other hand, if the machine learning unit 53 determines in step S120 not to continue the machine learning (NO), it stores the generated trained learning model 10 in the trained model storage unit 54 in step S130 and ends the machine learning.

[0064] (Device proposed by business partner) Next, a supplier proposal device 9 that proposes suppliers based on purchase specifications using the trained learning model 10 generated using the above-mentioned technique will be described.

[0065] The business partner proposal device 9 is configured with a general-purpose or dedicated computer. This computer may be a stationary computer or a portable computer, and may also be a client computer, a server computer, or a cloud computer. As shown in FIG. 6, the business partner proposal device 9 includes an input data acquisition unit 91, an inference unit 92, a trained model storage unit 93, and an output processing unit 94.

[0066] The input data acquisition unit 91 functions as an interface unit that acquires input data for estimating the reduction rate for the order target via the communication network 12. This input data acquisition unit 91 is connected via the communication network 12 to the product type information assignment unit 3, the supplier candidate extraction unit 8, the worker terminal 13 used by the worker, and the like.

[0067] The product type information assignment unit 3 automatically assigns product type information to the analysis information (LSA similarity) obtained by applying latent semantic analysis to the purchase specifications created for the item to be ordered, by referring to the product type information database 4.

[0068] The supplier candidate extraction unit 8 extracts selectable supplier candidates with reference to the supplier information database 7 based on the product type information assigned by the product type information assignment unit 3. Specifically, as shown in FIG. 7, the supplier information database 7 stores various information related to suppliers, such as the supplier's company name, the business items (handled items) listed on the company's website, and past purchase records. Therefore, the supplier candidate extraction unit 8 may extract suppliers whose business items (handled items) include the same product type based on the product type information assigned by the product type information assignment unit 3. Furthermore, suppliers with past purchase records may also be extracted.

[0069] The input data acquisition unit 91 acquires as input data various data consisting of the product type information assigned to the analysis information of the purchase specification α2 by the product type information assignment unit 3, the supplier information of the candidate suppliers extracted by the candidate supplier extraction unit 8, and the purchasing conditions presented by the purchasing condition presentation unit 6, or the purchasing conditions finally selected by a procurement department staff member based on the above.

[0070] Here, the purchase condition presentation unit 6 presents purchase conditions that match the product type assigned by the product type information assignment unit, for example, using a tree-structured learning model. Specifically, the method disclosed in JP 2022-91337 A is used, in which the purchase conditions and customer information described below are used as independent variables, and the reduction rate is used as a dependent variable, and a tree-structured learning model is created in advance for each product type. From the product type information assigned by the product type information assignment unit 3, a decision tree linked to it is identified, and the purchase conditions of the decision tree are presented. (Purchase conditions) Classification of competitive or special contract, contract method, budget price, contract month, payment terms, difference between quotation request date and delivery date, delivery destination, contract manager, technical manager, contract manager age (Business Partner Information) Contractor name, contract classification, number of years of business with contractor, etc.

[0071] In this example, the purchasing condition presentation unit 6 is connected to the input data acquisition unit 91 via the worker terminal 13, and the presented purchasing conditions are first transmitted to the worker terminal 13. The purchasing conditions presented by the purchasing condition presentation unit 6 are input directly to the input data acquisition unit 91 without being adjusted by the worker terminal 13, or, if necessary, are adjusted by the person in charge with reference to the purchasing conditions presented at the worker terminal 13, and then input to the input data acquisition unit 91 as the final selected purchasing conditions.

[0072] In the supplier proposal device 9, when the input data necessary for inference in the inference unit 92 is acquired by the input data acquisition unit 91, the inference unit 92 inputs the product type information acquired by the input data acquisition unit 91, the finally selected purchasing conditions, and supplier information of one supplier extracted from the supplier candidates into the learning model 10 stored in the trained model memory unit 93, and estimates the reduction rate of the extracted supplier.

[0073] This inference process uses a learned learning model 10 trained by the machine learning device 5. The inference unit 92 not only has a function of performing inference processing using the learning model 10, but may also include a preprocessing function of adjusting the input data acquired by the input data acquisition unit 91 into a desired format or the like and inputting the adjusted data into the learning model 10 as preprocessing of the inference process.

[0074] The trained model storage unit 93 is a database that stores trained learning models 10 used in the estimation process of the inference unit 92. The number of learning models 10 stored in the trained model storage unit 93 is not limited to one. For example, multiple learning models 10 in cases where the number of input data is different or where machine learning techniques are different may be stored and appropriately selected for use.

[0075] After the inference unit 92 estimates the reduction rate of each of the supplier candidates, the output processing unit 94 causes the display unit 11 to display the suppliers with the highest reduction rates as final supplier candidates in descending order of reduction rate. At this time, each of the final supplier candidates may be displayed together with the corresponding reduction rate, or the final supplier candidate may be displayed together with the largest reduction rate (expected value of reduction rate). The display unit 11 may be replaced by the monitor of the worker terminal 13.

[0076] (How to propose to business partners) FIG. 8 is a flowchart showing an example of a supplier suggestion method using supplier suggestion device 9. Hereinafter, the process up to suggesting final supplier candidates will be described with reference to this flowchart.

[0077] First, in step S200, when a purchasing specification for a new ordering target is obtained, the analysis information acquisition unit 2b performs latent semantic analysis on this purchasing specification α2 to acquire analysis information (LSA similarity) of the purchasing specification. The business partner proposal device 9 then refers to the product type information database 4 and assigns product type information corresponding to the ordering target to the analysis information (LSA similarity) of the purchasing specification for the new ordering target.

[0078] Thereafter, based on the assigned product type information or by judging the note information, it is determined whether or not the selection of a supplier is unnecessary and the party to which a quotation is requested can be determined, i.e., whether or not the selection of a supplier is necessary (step S210).

[0079] If it is determined that selection of a supplier is not necessary and that the supplier for quotation can be determined, for example, if the technical manager has registered a special request as note information, or if it has been determined based on the product type information that there is only one manufacturer available, or if it has been determined that orders will be placed at the beginning of the fiscal year to allocate the share of procurement sources, then there is no need for this device to consider a supplier, and the device will end the process without further processing.

[0080] On the other hand, if it is determined that the selection of a supplier is necessary, in step S220, selectable supplier candidates are extracted from the product type information assigned in step S200 by referring to the supplier information database 7 (step S220). The supplier candidates extracted at this stage are extracted as candidates from all suppliers with the possibility of doing business by referring to the items handled (business items) on the supplier's homepage, etc.

[0081] Thereafter, the purchasing condition presenting unit 6 presents purchasing conditions that match the currently assigned product type information from a decision tree that has been generated in advance for each product type information (S230).

[0082] The person in charge may use the presented purchasing conditions as they are, but may also refer to the presented purchasing conditions and change the purchasing conditions as necessary to select the final purchasing conditions. For example, even if the purchasing condition presentation unit suggests that the delivery date be extended, if the delivery date cannot be changed at that time, the person in charge changes the delivery date conditions and sets them as the final purchasing conditions.

[0083] Then, in step S240, the product type information assigned in step S200, the purchasing conditions finally determined by referring to the purchasing conditions presented in step S230, and the supplier information of the selectable supplier candidates extracted from the supplier information database 7 in step S220 are used as input data to estimate the reduction rate of each supplier candidate using the trained learning model 10.

[0084] Thereafter, in step S250, supplier candidates with the top predicted reduction rates are displayed as final supplier candidates.

[0085] 9 shows an example in which the support device displays final supplier candidates based on the specifications for purchasing equipment XX. Based on the product type information given to the analysis information of the specifications for purchasing equipment XX, the purchasing conditions presentation unit 6 proposes purchasing conditions in which competitive bidding is performed, the number of companies requested to make a quotation is set to three, and the contract method is set to the target price method. The procurement department staff refers to this and ultimately selects "competitive bidding", changes the number of companies requested to make a quotation to four, and selects the target price method as the contract method. As a result, the top four companies among the supplier candidates with the highest reduction rates are displayed together with the reduction rates, and the highest reduction rate predicted by the learning model is displayed as the reduction rate expected in competitive bidding (expected reduction rate). If the staff is satisfied with the displayed purchasing conditions and supplier candidates, they press the Confirm button to confirm the order settings in order to request a quotation from the displayed final supplier candidates.

[0086] Therefore, according to the above-mentioned ordering support device 1, it is possible to automatically obtain product type information of the order target by referring to a product type information database that associates the analysis information (LSA similarity) with product type information based on the analysis information (LSA similarity) obtained by performing latent semantic analysis on the items described in the purchase specifications, thereby making it possible to shorten the time required to identify the order target.

[0087] In addition, possible suppliers are extracted based on product type information, and a learning model that has been machine-learned to detect correlations between input data including product type information, supplier information, and purchasing conditions, and output data including discount rates, is used to estimate discount rates for each extracted supplier from the product type information of the order target, the final selected purchasing conditions that are finally selected by referring to the purchasing conditions presented by the purchasing condition presentation unit 6, and the supplier information of each of the extracted candidate suppliers, and the suppliers with the highest discount rates are displayed, so that purchasers, regardless of experience (from beginners to veterans), can effectively find suppliers that are likely to be able to purchase the order target at the lowest price.

[0088] In the above embodiment, a case has been described in which a neural network is adopted as a specific method of machine learning by the machine learning unit 53, but the machine learning unit 53 may adopt any other machine learning method. For example, tree types such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural net types (including deep learning) such as recurrent neural networks and convolutional neural networks, clustering types such as hierarchical clustering, non-hierarchical clustering, k-nearest neighbors and k-means, multivariate analyses such as principal component analysis, factor analysis, and logistic regression, support vector machines, and the like may be adopted.

[0089] In addition, the above-mentioned machine learning device 5 and business partner proposal device 9 can also be provided in the form of a program (machine learning program or business partner proposal program) for executing each step of the above-mentioned machine learning method and business partner proposal method. [Explanation of symbols]

[0090] 1 Order support device 2a, 2b Analysis information acquisition part 3 Product type information provision department 4. Product type information database 6 Purchasing conditions presentation department 7. Business Partner Information Database 8. Supplier candidate extraction section 10 Learning Model 54, 93 Trained model memory α1, α2 Purchase Specifications

Claims

1. An ordering support device that supports ordering of an order target, an analysis information acquisition unit that performs latent semantic analysis on a purchase specification created to identify an order target and acquires analysis information of the purchase specification; a product type information database that stores analysis information acquired by the analysis information acquisition unit for past purchase specifications and product type information assigned to an order target identified in the past purchase specifications in association with each other; A business partner information database for storing business partner information; a purchase condition presentation unit that presents purchase conditions that match the product type; a learning model storage unit that stores a learning model that has been machine-learned to learn a correlation between input data including the product type information, the customer information, and the purchasing conditions, and output data including a reduction rate that indicates a reduction rate from the budget amount to the contract amount; Equipped with A product type information assigning unit that assigns product type information to the new order target by referring to the product type information database based on analysis information obtained by performing latent semantic analysis on the purchase specification of the new order target by the analysis information acquisition unit; a supplier candidate extraction means for extracting selectable supplier candidates by referring to the supplier information database based on the product type information assigned by the product type information assignment unit; a reduction rate estimation means for estimating a reduction rate for each of the supplier candidates using the learning model from the product type information assigned by the product type information assignment unit, the purchasing conditions presented by the purchasing condition presentation unit for the product type to be newly ordered or final selected purchasing conditions finally selected with reference thereto, and supplier information of each of the supplier candidates extracted by the supplier candidate extraction means; a final supplier candidate display means for displaying the supplier candidates with the top reduction rates estimated by the reduction rate estimation means as final supplier candidates; An ordering support device comprising:

2. The purchase condition presentation unit, The ordering support device according to claim 1, characterized in that a tree-structured learning model is created for each product type, in which the purchasing conditions and the supplier information are independent variables and the reduction rate is a dependent variable, a decision tree linked to the product type information assigned by the product type information assigning unit is identified, and the purchasing conditions of the decision tree are presented.

3. an analysis information acquisition unit that performs latent semantic analysis on a purchase specification created to identify an order target and acquires analysis information of the purchase specification; a product type information database that stores analysis information acquired by the analysis information acquisition unit for past purchase specifications and product type information assigned to an order target identified in the past purchase specifications in association with each other; A business partner information database for storing business partner information; a purchase condition presentation unit that presents purchase conditions according to the type of product; a learning model storage unit that stores a learning model that has been machine-learned to learn a correlation between input data including the product type information, the customer information, and the purchasing conditions, and output data including a reduction rate that indicates a reduction rate from the budget amount to the contract amount; An ordering support method for supporting ordering of an order target using an ordering support device comprising: A product type information assignment step of assigning product type information to the new order target by referring to the product type information database based on analysis information obtained by performing latent semantic analysis on the purchase specification of the new order target by the analysis information acquisition unit; a supplier candidate extraction step of extracting selectable supplier candidates by referring to the supplier information database based on the product type information assigned in the product type information assignment step; a reduction rate estimation step of estimating a reduction rate for each of the supplier candidates using the learning model from the product type information assigned in the product type information assignment step, the purchasing conditions presented by the purchasing condition presentation unit for the product type to be newly ordered or final selected purchasing conditions finally selected with reference thereto, and supplier information of each of the supplier candidates extracted in the supplier candidate extraction step; a final supplier display step of displaying the supplier candidates with the top reduction rates estimated in the reduction rate estimation step as final supplier candidates; An ordering support method comprising the steps of:

4. An ordering support program for causing a computer to execute each step of the ordering support method according to claim 3.

Images