Cosmetics manufacturing support device, cosmetics manufacturing support method, and cosmetics manufacturing support program

Abstract

In order to suitably assist in new cosmetic product development, this cosmetics manufacturing assistance device (1) comprises an acceptance unit (11) for accepting a request relating to new cosmetic product development, an estimation unit (12) for estimating a new product manufacturing method conforming to the request on the basis of a trained model that is trained on the relationship between the components and efficacy of an existing product and a method for manufacturing the existing product, and an output unit (13) for outputting information indicating the estimated manufacturing method.

Description

[Technical Field]

[0001] The present invention relates to a cosmetics manufacturing support device and the like that generates information related to the development of cosmetics. [Background technology]

[0002] In developing new cosmetic products, the process involves selecting from a vast number of combinations of materials and manufacturing methods those that are deemed promising for realizing the product concept, producing a prototype product, and testing it to verify its effectiveness (see, for example, Patent Document 1 below). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2017-226612 Summary of the Invention [Problem to be solved by the invention]

[0004] Developing new cosmetic products requires the above process to be repeated until a combination that realizes the product concept is found, which requires a great deal of effort and time. Therefore, there is a need for technology that can effectively support the development of new cosmetic products.

[0005] One aspect of the present invention has been made in consideration of the above-mentioned problems, and one of its objectives is to provide a technology that favorably supports the development of new cosmetic products. [Means for solving the problem]

[0006] A cosmetics manufacturing support device according to one aspect of the present invention includes a receiving means for receiving requests for the development of new cosmetic products, a trained model that has learned the relationship between the ingredients and effects of existing cosmetic products and manufacturing methods for the existing products, an estimation means for estimating a manufacturing method for a new product that matches the request based on the request, and an output means for outputting information indicating the manufacturing method estimated by the estimation means.

[0007] In one aspect of the present invention, a computer receives a request for the development of a new cosmetic product, estimates a manufacturing method for the new product that matches the request based on a trained model that has learned the relationship between the ingredients and efficacy of existing cosmetic products and the manufacturing methods of the existing products, and the request, and outputs information indicating the estimated manufacturing method.

[0008] A cosmetics development support program according to one aspect of the present invention causes a computer to perform the following processes: accepting a request for the development of a new cosmetic product; estimating a manufacturing method for a new product that matches the request based on a trained model that has learned the relationship between the ingredients and efficacy of an existing cosmetic product and a manufacturing method for the existing product, and the request; and outputting information indicating the estimated manufacturing method. [Effects of the Invention]

[0009] According to one aspect of the present invention, it is possible to preferably support the development of new cosmetic products. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram showing the configuration of a cosmetics manufacturing support device according to a first exemplary embodiment of the present invention. [Figure 2] 1 is a flow chart showing the flow of a cosmetics manufacturing support method according to a first exemplary embodiment of the present invention. [Figure 3] FIG. 1 is a diagram illustrating feature learning in graph-based relational learning. [Figure 4] FIG. 10 is a block diagram showing the configuration of a cosmetics manufacturing support device according to a second exemplary embodiment of the present invention. [Figure 5] FIG. 10 is a diagram illustrating an example of calculation of a recommendation level by an evaluation unit. [Figure 6] FIG. 10 is a diagram showing an example of a presentation of existing products in response to a request. [Figure 7] 10A and 10B are diagrams illustrating a method for identifying similar products by a link prediction unit. [Figure 8] FIG. 10 is a diagram showing an example of presentation of evidence information based on the prediction result of the probability that a predetermined property will appear in a new product. [Figure 9] FIG. 10 is a flowchart showing the flow of processing executed by a cosmetics manufacturing support device according to a second exemplary embodiment of the present invention. [Figure 10] FIG. 10 is a diagram showing an overview of a cosmetics manufacturing support method according to a third exemplary embodiment of the present invention. [Figure 11] FIG. 10 is a block diagram showing the configuration of a cosmetics manufacturing support device according to a third exemplary embodiment of the present invention. [Figure 12] FIG. 10 is a flowchart showing the flow of processing executed by a cosmetics manufacturing support device according to a third exemplary embodiment of the present invention. [Figure 13] FIG. 10 is a diagram showing an overview of a cosmetics manufacturing support method according to a fourth exemplary embodiment of the present invention. [Figure 14] FIG. 10 is a block diagram showing the configuration of a cosmetics manufacturing support device according to a fourth exemplary embodiment of the present invention. [Figure 15] FIG. 10 is a flowchart showing the flow of processing executed by a cosmetics manufacturing support device according to a fourth exemplary embodiment of the present invention. [Figure 16] FIG. 10 is a diagram showing an overview of a cosmetics manufacturing support method according to a fifth exemplary embodiment of the present invention. [Figure 17] FIG. 10 is a flowchart showing the flow of processing executed by a cosmetics manufacturing support device according to a fifth exemplary embodiment of the present invention. [Figure 18] FIG. 10 is a diagram illustrating an example of predicting the properties of a new product based on feature amounts calculated from a new product graph and an existing product graph. [Figure 19]FIG. 1 is a configuration diagram for realizing a cosmetics manufacturing support device using software. DETAILED DESCRIPTION OF THE INVENTION

[0011] Exemplary Embodiment 1 A first exemplary embodiment of the present invention will be described in detail with reference to the drawings. This exemplary embodiment is a basic form of the exemplary embodiments described below.

[0012] (Cosmetics manufacturing support equipment) The configuration of a cosmetics manufacturing support device 1 according to this exemplary embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the cosmetics manufacturing support device 1. As shown in the figure, the cosmetics manufacturing support device 1 includes a reception unit (reception means) 11, an estimation unit (estimation means) 12, and an output unit (output means) 13.

[0013] The receiving unit 11 receives a request for development of a new cosmetic product. The estimation unit 12 estimates a manufacturing method for the new product that matches the request, based on a trained model that has learned the relationship between the ingredients and effects of existing cosmetic products and the manufacturing methods of the existing products, and the request received by the receiving unit 11. The output unit 13 outputs information indicating the manufacturing method estimated by the estimation unit 12.

[0014] The "manufacturing method" may include, for example, the blending ratio of ingredients, the processing method of raw materials (pre-processing method, mixing time, mixing speed, post-processing method, etc.), the facilities and equipment used in the production, etc. Also, the ingredients of a new product may be estimated as the "manufacturing method."

[0015] The cosmetics manufacturing support device 1 having the above configuration can present to the user a manufacturing method for a new product that is deemed to meet the user's request, based on the relationship between the ingredients and effects of existing cosmetics and the manufacturing method for the existing product. Therefore, the above configuration has the effect of providing optimal support for the development of new cosmetic products.

[0016] (Cosmetics Manufacturing Support Program) The functions of the above-described cosmetics manufacturing support device 1 can also be realized by a program. The cosmetics manufacturing support program according to this exemplary embodiment causes a computer to execute the following processes: accepting a request for the development of a new cosmetic product; estimating a manufacturing method for a new product that matches the request based on the request and a trained model that has learned the relationship between the ingredients and efficacy of existing cosmetic products and manufacturing methods of the existing products; and outputting information indicating the estimated manufacturing method. This cosmetics manufacturing support program provides the effect of optimally supporting the development of new cosmetic products.

[0017] (Cosmetics manufacturing support method) A cosmetics manufacturing support method according to this exemplary embodiment will be described with reference to Fig. 2. Fig. 2 is a flow chart showing the flow of a cosmetics manufacturing support method according to a first exemplary embodiment of the present invention.

[0018] In S11, the computer receives a request for the development of a new cosmetic product. The request may be received via any input device. For example, the request may be received via a mouse, keyboard, touch panel, or voice input device.

[0019] In S12, the computer estimates a manufacturing method for a new product that matches the request received in S11, based on a trained model that has learned the relationship between the ingredients and efficacy of existing cosmetic products and the manufacturing methods of the existing products.

[0020] In S13, the computer outputs information indicating the estimated manufacturing method to any device, such as a display device to display the information, or an audio output device to output the information as audio.

[0021] As described above, in the cosmetics manufacturing support method according to this exemplary embodiment, a computer receives a request for the development of a new cosmetic product (S11), estimates a manufacturing method for the new product that matches the request based on the request received in S11 and a trained model that has learned the relationship between the ingredients and efficacy of existing cosmetic products and manufacturing methods for the existing products (S12), and outputs information indicating the manufacturing method estimated in S12 (S13). This cosmetics manufacturing support method has the effect of providing optimal support for the development of new cosmetic products.

[0022] Note that each step in the above-described cosmetics manufacturing support method may be executed by a single computer (e.g., cosmetics manufacturing support device 1), or each step may be executed by a different computer. This also applies to the flows described in exemplary embodiment 2 and subsequent embodiments.

[0023] [Graphs and Learning] Below, we will explain graphs, which are an example of information that can be used to support the development of cosmetics, in Exemplary Embodiment 1 and the following Exemplary Embodiments (hereinafter referred to as each Exemplary Embodiment). We will also explain learning of the graphs and predictions using the graphs.

[0024] (graph) A graph here refers to data with a structure consisting of multiple nodes and links connecting the nodes. The type of link that represents the relationship between nodes is also called a "relation." Links are also sometimes called edges. Graphs can be broadly divided into directed graphs, in which each link has a direction, and undirected graphs, in which each link does not have a direction. It is possible to use either directed graphs or undirected graphs, or to use a combination of the two.

[0025] In each exemplary embodiment, when a graph is used, the nodes may represent tangible or intangible elements related to the manufacturing method of the cosmetic product. For example, Ingredients, ingredient composition, raw materials, materials Product Category Development concept Properties (efficacy, scent, appearance, side effects, etc.) Product identification information such as product name and product ID Manufacturing method (manufacturing equipment used, manufacturing process, manufacturing conditions, etc.) It is possible to use a graph including nodes representing various elements such as the above. Note that the graph may include multiple nodes corresponding to one element. For example, as nodes representing the ingredients of a cosmetic product, each ingredient contained in the cosmetic product may be represented by an individual node, such as a node representing a first ingredient, a node representing a second ingredient, and a node representing a third ingredient. The same applies to other elements.

[0026] If there are nodes corresponding to the elements above, the links connecting such nodes are ·Relationship between certain elements and components Relationship between an element and a product category -Relationship between certain elements and development concept Relationship between certain elements and properties The relationship between an element and a product For example, a link connecting a node indicating an ingredient with a node indicating an effect may represent a relationship in which the ingredient is a factor in the effect.

[0027] (Learning and Prediction) For the graphs described above, graph-based relationship learning can be performed by applying machine learning techniques. Such learning enables classification and prediction processes to be performed using the graphs. Note that in each exemplary embodiment, such learning may be performed as part of support for the manufacture of cosmetics, or a trained graph that has already undergone such learning may be used.

[0028] In graph-based relationship learning, first, the feature of each node is calculated. The feature may be in vector format, for example. By expressing the feature of each node as a feature vector, it becomes possible to learn about graphs in which nodes of various formats coexist. For example, graph-based relationship learning can also be performed on graphs that include images, numerical values, and the like that indicate the various elements described above.

[0029] Next, the feature values ​​of each node are updated based on the links connected to each node and the nodes to which those links are connected. This process is similar to the convolution process in a convolutional neural network. This will be explained with reference to Figure 3. Figure 3 is a diagram explaining feature learning in graph-based relational learning.

[0030] The graph shown in Fig. 3 includes four nodes A to D. Node A is connected to nodes B and C, and node C is connected to node D. After calculating the initial features of these four nodes, convolution is performed multiple times as described below to update the features of each node.

[0031] In the first convolution, the initial feature of node A is multiplied by a predetermined weight and added to the feature of nodes B and C connected to node A. For node C, the initial feature of node C is multiplied by a predetermined weight and added to the feature of node D. If it is a directed graph, the weight is adjusted according to the direction of the link.

[0032] In the second convolution, as in the first convolution, the feature of each node is multiplied by a predetermined weight and then added with the feature of the node linked to that node. Here, the feature of node C reflects the feature of node D due to the first convolution. Therefore, in the second convolution, not only the feature of node C but also the feature of node D is reflected in node A.

[0033] By repeating the above process a number of times according to the node hierarchy, the features of each node directly or indirectly connected by links are mutually reflected. In graph-based relationship learning, the weight values ​​used for the above weighting are optimized based on the known relationships between nodes. By using such a trained graph (which can also be called a trained model), it is possible to predict the relationships between nodes and the nodes to which links will lead, as described below.

[0034] (Inter-node relationship prediction) By performing the above-described learning, it becomes possible to predict relationships between nodes that are not explicitly stated in the original graph. To perform node-to-node relationship prediction, a user simply specifies two nodes and requests that the relationship between those nodes be returned. For example, if a user inputs a request asking about the relationship between the "Product A" node and the "Moisturizing" node, node-to-node relationship prediction can predict that the relationship, i.e., the link connecting these nodes, is "Efficacy." Furthermore, node-to-node relationship prediction can also calculate the probability (likelihood) of the predicted result. The same applies to node prediction, which will be described below.

[0035] (node ​​prediction) Furthermore, by performing the above-described learning, it is also possible to predict the nodes connected to a certain node via a predetermined link. To perform node prediction, a user simply specifies one node and a link starting from that node, and requests that the node at the link destination be returned. For example, suppose a user inputs a request asking about the nodes connected to the node for "Product A" via the "Effect" link. In this case, node prediction can predict whether the node connected to the node for "Product A" via the "Effect" link is "moisturizing" or "prevents dryness," for example.

[0036] Exemplary Embodiment 2 (Device configuration) The configuration of a cosmetics manufacturing support device 2 according to a second exemplary embodiment of the present invention will be described with reference to Fig. 4. Fig. 4 is a block diagram showing the configuration of a cosmetics manufacturing support device 2 according to this exemplary embodiment.

[0037] As shown in the figure, the cosmetics manufacturing support device 2 includes a reception unit 201, an existing product identification unit 202, an evaluation unit 203, a graph generation unit 204, a link prediction unit 205, a learning unit 206, a property prediction unit 207, an estimation unit 208, a basis generation unit 209, and an output unit 210.

[0038] In addition to these components, cosmetics manufacturing support device 2 may also include an input device that accepts user input operations, an output device that outputs data from cosmetics manufacturing support device 2, and a communication device that enables cosmetics manufacturing support device 2 to communicate with other devices. The output mode of the output device is arbitrary, and may be, for example, a display output or an audio output.

[0039] The receiving unit 201 receives requests for the development of new cosmetic products. The requests may include, for example, the efficacy required for the new product, the raw materials or ingredients to be used in the new product, the manufacturing method to be used when manufacturing the new product, etc.

[0040] The existing product identification unit 202 identifies existing products having properties that match the request received by the reception unit 201, based on the nodes and links included in the existing product graph. Note that properties refer to properties or states, and include, for example, efficacy, the presence or absence of side effects, the details of the side effects, texture, whether the product is liquid or solid, and so on.

[0041] An existing product graph is a graph that represents one or more existing products with nodes that represent the ingredients, efficacy, or manufacturing method of the existing product, and edges that represent the relationships between the nodes. An existing product graph is a trained graph of the relationships between the nodes, and is a trained model. An existing product graph can also be called a knowledge graph. Note that a graph corresponding to one existing product may be called an existing product graph, or a graph corresponding to multiple existing products may be called an existing product graph.

[0042] The evaluation unit 203 evaluates the existing products identified by the existing product identification unit 202 and calculates their recommendation levels. Various evaluation criteria can be applied. For example, evaluation may be based on the degree of suitability for the request. For example, assume that the request includes a desired effect of the new product. In this case, the evaluation unit 203 may calculate the recommendation level so that the recommendation level of an existing product corresponding to an existing product graph including a node indicating that effect is higher than the recommendation level of an existing product corresponding to an existing product graph not including a node indicating that effect.

[0043] The graph generation unit 204 generates a new product graph that represents a new product that the user wants to manufacture based on information about the new product. Specifically, the graph generation unit 204 generates a new product graph that represents the new product that the user wants to manufacture with nodes that represent the ingredients, efficacy, or manufacturing method of the new product, and edges that represent the relationships between the nodes. Note that the information about the new product that the user wants to manufacture is included in the request received by the receiving unit 201.

[0044] The link prediction unit 205 uses the new product graph and the existing product graph to perform link prediction to predict the relationship between nodes that are not connected by links in the new product graph and the existing product graph, thereby identifying existing products that are similar to the new product. Note that, hereinafter, existing products that are similar to the new product will be referred to as similar products.

[0045] The learning unit 206 learns the relationships between the nodes included in the existing product graph based on various information about existing products, and generates a learned existing product graph. Unless otherwise specified, the existing product graph refers to one that has been learned by the learning unit 206. The learned existing product graph may also be loaded into the cosmetics manufacturing support device 2, in which case the learning unit 206 may be omitted.

[0046] The property prediction unit 207 predicts the properties of the new product. Specifically, the property prediction unit 207 identifies the properties shown in the existing product graph by referring to the existing product graph of the similar product identified by the link prediction unit 205. Then, the property prediction unit 207 sets the properties identified in this way as the predicted results of the properties of the new product.

[0047] The estimation unit 208 estimates a manufacturing method for a new product that matches the request received by the reception unit 201, based on a trained model that has learned the relationship between the ingredients and efficacy of existing products, which are existing cosmetic products, and the manufacturing methods of the existing products, and the request received by the reception unit 201.

[0048] As described above, the trained model may be an existing product graph in which one or more existing products are represented by nodes representing the ingredients, efficacy, or manufacturing methods of the existing products, and edges representing the relationships between the nodes. With this configuration, it is possible to estimate an appropriate manufacturing method by taking into account the mutual relationships between the ingredients, efficacy, and manufacturing methods of the existing products.

[0049] The basis generating unit 209 generates basis information indicating the basis of the estimation by the estimating unit 208. Various methods can be applied to generate the basis information. The method of generating the basis information will be described later.

[0050] As described above, output unit 210 outputs various information generated by cosmetics manufacturing support apparatus 2, such as information indicating the manufacturing method estimated by estimation unit 208. The destination of the information is arbitrary; for example, if cosmetics manufacturing support apparatus 2 is equipped with an output device as described above, the information may be output to that output device. Alternatively, the information may be output to an output device external to cosmetics manufacturing support apparatus 2, for example.

[0051] As described above, the request received by the receiving unit 201 may include information indicating the efficacy desired for the new product. In this case, the estimation unit 208 may estimate, as a manufacturing method for manufacturing the new product having the efficacy, a manufacturing method that includes some elements of the manufacturing method for the existing product corresponding to the node and link indicating the efficacy included in the existing product graph.

[0052] The elements included in the manufacturing method of the existing product corresponding to the nodes and links indicating the efficacy desired for the new product may be factors that cause the efficacy to be realized. Therefore, with the above configuration, it is possible to estimate the manufacturing method for a new cosmetic product with the desired efficacy.

[0053] As described above, the cosmetics manufacturing support device 2 is also equipped with a link prediction unit 205 that identifies existing products that have a predetermined relationship with the new product by link prediction using a new product graph that includes multiple nodes related to the manufacturing of the new product to be developed and an existing product graph to predict the relationship between nodes that are not connected by links in the new product graph and the existing product graph, and the estimation unit 208 estimates a manufacturing method for the new product that matches the request, based on the manufacturing method for the existing product identified by the link prediction unit 205.

[0054] The manufacturing method of an existing product that has a predetermined relationship with the new product to be developed is useful information for estimating a manufacturing method of a new product that meets the request. With the above configuration, it is possible to estimate a manufacturing method of a new product that meets the request by taking this useful information into consideration.

[0055] As described above, the cosmetics manufacturing support device 2 also includes an existing product identification unit 202 that identifies existing products having properties that match the request received by the reception unit 201 based on the nodes and links included in the existing product graph.

[0056] Existing products having properties that match the request are useful information for estimating a manufacturing method for a new product that matches the request. Therefore, by providing the existing product identification unit 202, it is possible to estimate a manufacturing method for a new product that matches the request by taking this useful information into consideration. For example, information indicating the existing products identified by the existing product identification unit 202 may be output to the output unit 210. This allows the user to recognize what existing products have properties that match the request.

[0057] (How to calculate the recommendation level) Fig. 5 is a diagram showing an example of calculation of the recommendation level by the evaluation unit 203. In the example of Fig. 5, the recommendation level is calculated for each of existing products A and D. Note that in this example, it is assumed that the request received by the receiving unit 201 specifies two effects, "prevents rough skin" and "prevents dryness," as effects required of a new product. In this example, it is also assumed that the recommendation level is set to "5" when one effect matches the request.

[0058] The existing product graph for existing product A includes a node indicating that existing product A has the effect of "preventing rough skin." Therefore, the evaluation unit 203 calculates the recommendation level for existing product A as "5." On the other hand, the existing product graph for existing product D includes a node indicating that existing product D has the effect of "preventing rough skin" and a node indicating that existing product D has the effect of "preventing dryness." Therefore, the evaluation unit 203 calculates the recommendation level for existing product D as "10." In this way, by determining in advance the degree of suitability for a request and the recommendation level corresponding to that degree, the evaluation unit 203 can calculate the recommendation level for each existing product.

[0059] Furthermore, the evaluation unit 203 may increase the recommendation level of an existing product that is more similar to the new product. For example, assume that the new product graph includes nodes and links that indicate multiple ingredients of the new product and their blending ratios. In this case, the recommendation level of an existing product that shares ingredients with the new product may be increased compared to an existing product that does not share ingredients with the new product, and the recommendation level of an existing product that is closer to the blending ratio of the new product may be increased.

[0060] In addition, the evaluation unit 203 may, for example, give a higher recommendation level to an existing product that can be manufactured with manufacturing equipment available to the user than to an existing product that cannot be manufactured with manufacturing equipment available to the user. The manufacturing equipment available to the user may be input as a request. The evaluation unit 203 may also calculate the recommendation level by combining the various evaluation criteria described above.

[0061] (Example of an existing product presented in response to a request) The output unit 210 may output the existing products identified by the existing product identification unit 202 in response to the request received by the reception unit 201 and the recommendation levels calculated by the evaluation unit 203 in a format such as that shown in Fig. 6. Fig. 6 is a diagram showing an example of how existing products are presented in response to a request.

[0062] In the example of Figure 6, the product name, ingredients, manufacturing method, efficacy (or side effects), and recommendation level of the existing products identified by the existing product identification unit 202 are presented. In addition, the existing products are presented by category. Such various information can be extracted from the existing product graph. By presenting the information shown in Figure 6 to the user, the user can infer the ingredients and manufacturing method that will give the new product the desired efficacy.

[0063] (Regarding identification of similar products) Fig. 7 is a diagram illustrating a method for identifying similar products by the link prediction unit 205. Fig. 7 shows an existing product graph in which existing products A to C are represented by nodes that represent the ingredients, efficacy, or manufacturing method of the existing products, and edges that represent the relationships between the nodes. For example, the node for "existing product A" and the node for "prevents rough skin" are connected by an edge that indicates "efficacy," which indicates that existing product A has the efficacy of preventing rough skin.

[0064] 7 includes nodes for multiple existing products, existing products A to C, but a graph consisting of only nodes related to one existing product may also be called an existing product graph. In this case, similar products are identified using multiple existing product graphs corresponding to each of the multiple existing products.

[0065] Such an existing product graph can be generated from the ingredients, effects, and manufacturing methods of each existing product. Furthermore, by learning the relationship between the ingredients and effects shown in the existing product graph and the manufacturing methods of existing products, it becomes possible to infer the relationship between the ingredients and effects of a new product and the manufacturing methods of existing products.

[0066] 7 also shows a new product graph in which new products are represented by nodes that represent the ingredients, efficacy, or manufacturing method of the new product, and edges that represent the relationships between the nodes. More specifically, the new product graph shown in FIG. 7 includes nodes and links that indicate that the new product has the efficacy of "preventing rough skin," and nodes and links that indicate that the new product contains an ingredient called "ingredient x."

[0067] The graph generating unit 204 generates the new product graph as described above based on the request for a new product received by the receiving unit 201. That is, when the efficacy required of the new product is to prevent rough skin and the request to use ingredient x in the new product is received, a new product graph as shown in Fig. 7 is generated.

[0068] By using the new product graph and existing product graph generated in this manner, the link prediction unit 205 can identify existing products that have a predetermined relationship with the new product. For example, it is possible to identify existing products that are similar to the new product, i.e., the similar products described above. It is also possible to identify existing products that are dissimilar to the new product, existing products that belong to the same category as the new product, existing products that share effects with the new product, etc. In this exemplary embodiment, an example of identifying similar products is described, but existing products that have other relationships with the new product may also be identified. Such identification results, like the identification results of similar products, can be used to estimate the manufacturing method of the new product.

[0069] The above identification can be achieved by link prediction, which predicts the relationship between nodes that are not connected by links in the new product graph and the existing product graph. For example, as shown by the dashed line in Figure 7, there is no link between the "new product" node in the new product graph and the "existing product A" node in the existing product graph.

[0070] By performing link prediction, the link prediction unit 205 can predict the probability that the relationship between these nodes is "similar." Similarly, the link prediction unit 205 can predict the probability that the relationship between the "new product" node and the nodes of existing products B and C included in the existing product graph is "similar." Then, the link prediction unit 205 can identify similar products based on the predicted probabilities. For example, the link prediction unit 205 may identify existing products whose predicted probability values ​​are equal to or greater than a threshold as similar products.

[0071] The link prediction unit 205 can also identify existing products that meet preset conditions or conditions set by the user as existing products that have a predetermined relationship with the new product. For example, it is possible to identify existing products that share at least some of the ingredients with the new product as similar products, or to identify existing products that share at least some of the manufacturing method with the new product as similar products.

[0072] (Method of generating evidence information) A method for generating evidence information by the evidence generation unit 209 will now be described. As described above, various methods can be applied to generate evidence information. For example, if the request received by the receiving unit 201 includes information indicating the efficacy desired for the new product, the evidence generation unit 209 may generate, as evidence for estimation, evidence information including information about existing products that have the same efficacy as the efficacy indicated in the request, i.e., the efficacy desired for the new product. The output unit 210 may then output the generated evidence information.

[0073] Since the existing product graph of an existing product that has the same efficacy as the efficacy desired for the new product includes a node indicating that efficacy, when generating the above-described evidence information, the evidence generation unit 209 first identifies the existing product graph that includes the node indicating that efficacy. Then, the evidence generation unit 209 generates evidence information based on the identified existing product graph. For example, the evidence generation unit 209 may generate evidence information that indicates the name and efficacy of the existing product shown in the identified existing product graph.

[0074] Information about existing products with the same efficacy as the efficacy desired for the new product is useful information for determining whether the new product manufactured using the estimated manufacturing method will have that efficacy. Therefore, the above configuration provides the user with useful information for determining whether the estimated manufacturing method is effective for imparting the desired efficacy to the new product.

[0075] Furthermore, the evidence information generated by the evidence generation unit 209 may include information indicating a manufacturing method for an existing product that has the same efficacy as the efficacy desired for the new product. The existing product graph of an existing product that has the same efficacy as the efficacy desired for the new product includes a node indicating that efficacy. Therefore, the evidence generation unit 209 first identifies an existing product graph that includes a node indicating that efficacy. Then, the evidence generation unit 209 generates evidence information indicating a manufacturing method for the existing product that is indicated in the identified existing product graph. For example, the evidence generation unit 209 may generate evidence information indicating at least one of the ingredients and manufacturing method indicated in the identified existing product graph.

[0076] The manufacturing method of an existing product that has the same efficacy as the efficacy desired for the new product is a valuable criterion for determining whether a new product manufactured by the estimated manufacturing method will have that efficacy. Therefore, the above configuration provides the user with valuable criterion for determining whether the estimated manufacturing method is effective for imparting the desired efficacy to the new product.

[0077] For example, when a manufacturing method for a new product having efficacy x is requested, the estimation unit 208 estimates the ingredients and manufacturing method of the new product having efficacy x. Here, it is assumed that the estimation unit 208 estimates a manufacturing method in which ingredients A and B are blended into the new product. In this case, the evidence information generation unit 209 may generate evidence information indicating that ingredient A is contained in existing product X1 having efficacy x, and that ingredient B is contained in existing product X2 having the same efficacy x. In this case, the evidence information generation unit 209 may also generate evidence information indicating a manufacturing method for existing product X1 (a manufacturing method including a step of blending ingredient A) or a manufacturing method for existing product X1 (a manufacturing method including a step of blending ingredient A).

[0078] Furthermore, the evidence generation unit 209 may generate evidence information based on the result of link prediction by the link prediction unit 205. In this case, the link prediction unit 205 uses the existing product graph and the new product graph, which include nodes indicating the effects required of the new product, to predict the probability that a node indicating the effect will be linked to a node included in the new product graph. Then, the evidence generation unit 209 generates evidence information according to the predicted probability.

[0079] This enables the basis generating unit 209 to generate basis information as shown in Fig. 8. Fig. 8 is a diagram showing an example of presentation of basis information based on the prediction result of the probability that a predetermined property will appear in a new product.

[0080] The evidence information shown in Fig. 8 includes the properties that the new product is expected to have and information indicating the likelihood that the new product will have those properties. Specifically, the evidence information shown in Fig. 8 includes the requested properties, i.e., the properties (e.g., efficacy) indicated in the request for a new product received by the receiving unit 201, and information indicating the likelihood that the new product will have those properties. Note that the relationship between the range of probability values ​​predicted by the link prediction unit 205 and the display of the likelihood of occurrence may be specified in advance. For example, if the predicted probability value is 90% or higher, it may be determined that the likelihood of occurrence is high.

[0081] The evidence information shown in Figure 8 also includes information indicating whether or not the new product has side effects and other properties related to the new product, and is associated with information indicating the possibility of these properties occurring.

[0082] By outputting the evidence information as shown in FIG. 8 by the output unit 210, the user can be provided with useful information for determining whether the estimated manufacturing method is effective for imparting the desired properties to the new product.

[0083] (Generating evidence for link prediction results) The basis generating unit 209 can also generate basis information by analyzing the new product graph and the existing product graph. A method for generating basis information by analyzing the new product graph and the existing product graph will be described below.

[0084] For example, the basis generator 209 may mine one or more rules from the new product graph and the existing product graph using PCA (Principal Component Analysis) reliability based on the OWA (Open-world assumption).The basis generator 209 may then generate basis information using the mined one or more rules.For example, the method described in the following document can be applied to rule mining.

[0085] Luis Galarraga et. al, “Fast rule mining in ontological knowledge bases with AMIE +”, The VLDB Journal(2015)24:707-730 As an example, the rule to be processed by the basis generating unit 209 is expressed as follows, using Head r(x, y) and Body { B1 , . . . , Bn}:

number

number

[0086] The basis generation unit 209 sets the following as the conditions for the mining process: Connected: All values ​​(variables, entities) in the rule are shared between different atoms. Closed: All variables in the rule must appear at least twice. Not reflexive: Rules containing reflexive atoms such as r(x, x) are not mined. Mining processing is carried out under the following conditions.

[0087] Furthermore, the basis generating unit 209

number

number

[0088] For example, suppose that the basis generation unit 209 has mined a rule that, for two products that satisfy the conditions of "common classification and common principal component nodes," "an element included in one manufacturing method can be applied to the manufacturing method of the other." In this case, when the link prediction unit 205 predicts that an element included in the manufacturing method of an existing product will be an element included in the manufacturing method of a new product, the basis generation unit 209 may generate, as the basis for this prediction, basis information indicating that the existing product and the new product "common classification and common principal component nodes."

[0089] (Processing flow) The flow of the process (cosmetics manufacturing support method) executed by cosmetics manufacturing support device 2 will be described with reference to Fig. 9. Fig. 9 is a flowchart showing the flow of the process executed by cosmetics manufacturing support device 2.

[0090] In S201, the receiving unit 201 receives a request for development of a new cosmetic product. In S201, for example, a request indicating the properties required for the new product, some of the raw materials to be used in manufacturing the new product, etc. is received.

[0091] In S202, the existing product identification unit 202 identifies an existing product having properties that match the request received by the reception unit 201, based on the nodes and links included in the existing product graph. For example, if the request received in S201 includes information indicating properties required for a new product, the existing product identification unit 202 detects nodes and links that indicate those properties from the existing product graph, and identifies an existing product that corresponds to the detected nodes and links.

[0092] In S203, the evaluation unit 203 calculates the recommendation level of the existing product identified in S202. If multiple existing products are identified in S202, the evaluation unit 203 calculates the recommendation level for each of the identified existing products.

[0093] In S204, the output unit 210 outputs the existing products identified in S202 and the recommendation levels calculated for the existing products in S203, thereby presenting this information to the user. For example, the output unit 210 may present the existing products and their recommendation levels to the user in a format as shown in FIG.

[0094] In S205, the receiving unit 201 receives input of information about a new product that the user wants to manufacture. The information about the new product may be information that allows a new product graph to be generated in the next S206. For example, the information about the new product may include information indicating the efficacy required for the new product, ingredients that have been decided to be used in the new product, etc.

[0095] In S205, the user can input the ingredients, blend amounts, and manufacturing method of the new product by referring to the information presented in S204. For example, the user can refer to the information shown in Fig. 6 to identify the ingredients and manufacturing method used in an existing product that has the efficacy desired for the new product, and input them as the ingredients and manufacturing method to be applied to the new product.

[0096] In S206, the graph generation unit 204 generates a new product graph based on the information input in S205. For example, if input of the efficacy required for the new product and the ingredients to be used in the new product is received in S205, the graph generation unit 204 may generate a new product graph including nodes and links indicating the efficacy and ingredients.

[0097] In S207, the link prediction unit 205 performs link prediction using the new product graph generated in S206 and the pre-trained existing product graph, and identifies similar products that are existing products similar to the new product. The link prediction unit 205 may identify multiple similar products. Here, the basis generation unit 209 may generate basis information that indicates the basis for the identification result of the link prediction unit 205 by analyzing the new product graph and the existing product graph.

[0098] In S208, the property prediction unit 207 predicts the properties of the new product. Specifically, the property prediction unit 207 predicts the properties of the similar products (identified in S207) shown in the existing product graph as the properties of the new product. Note that if multiple similar products are identified in S207, the property of each similar product is predicted in S208.

[0099] In S209, the estimation unit 208 estimates a manufacturing method for a new product that matches the request based on the pre-trained existing product graph and the request received in S201. More specifically, the estimation unit 208 first identifies a manufacturing method for the similar product identified in S207 by referring to the existing product graph. Next, the estimation unit 208 extracts some elements of the identified manufacturing method. In this case, the estimation unit 208 extracts some elements of the manufacturing method for a similar product whose properties predicted in S208 match the request. This makes it possible to extract elements that may have contributed to the similar product matching the request. Then, the estimation unit 208 estimates that the manufacturing method including the extracted elements is the manufacturing method for the new product that matches the request.

[0100] In S210, the basis generation unit 209 generates basis information indicating the basis for the estimation in S209. For example, the basis generation unit 209 may generate basis information including information on an existing product that has the same efficacy as the efficacy indicated in the request accepted in S201, i.e., the efficacy desired for the new product, as the basis for the estimation. Furthermore, this basis information may include information indicating a manufacturing method for the existing product that has the same efficacy as the efficacy desired for the new product.

[0101] Exemplary Embodiment 3 (overview) FIG. 10 is a diagram illustrating an overview of a cosmetics manufacturing support method according to this exemplary embodiment. In this exemplary embodiment, an example of supporting cosmetics manufacturing support using a base product graph and an existing product graph is described. Note that FIG. 10 shows nodes such as existing product A and existing product B as nodes included in the existing product graph, but other nodes included in the existing product graph are not shown. The existing product graph used in this exemplary embodiment is the same as that used in exemplary embodiment 2 (see FIG. 7).

[0102] A base product graph is a graph containing multiple nodes related to the production of a base product that will be the basis for a new product to be developed. The base product graph shown in Figure 10 contains a node indicating that the base product has the effect of "preventing rough skin," a node indicating that it contains an ingredient called "ingredient y," and a node and link indicating that it is produced in "mixing time y." Such a base product graph can be generated from the ingredients, effects, and production method of the base product.

[0103] For example, a base product graph can be generated by accepting input of the ingredients, efficacy, and manufacturing method of a base product as a user request. At this time, a request for efficacy, etc., required for a new product based on the base product may also be accepted. Alternatively, the user may select a base product from the existing products shown in the existing product graph. In this case, the existing product graph of the selected existing product can be used as the base product graph.

[0104] In the cosmetics manufacturing support method according to this exemplary embodiment, a base product graph and an existing product graph are used to identify existing products that have a predetermined relationship with the base product through link prediction for predicting relationships between nodes that are not connected by links in the base product graph and the existing product graph. Then, based on the manufacturing method of the identified existing product, a manufacturing method for a new product that is based on the base product and matches the request is estimated.

[0105] The manufacturing method of an existing product that has a predetermined relationship with the base product is useful information for estimating a manufacturing method of a new product that is based on the base product and matches the request. Therefore, the cosmetics manufacturing support method according to this exemplary embodiment can estimate a manufacturing method of a new product that matches the request by taking this useful information into consideration.

[0106] Link prediction makes it possible to identify existing products that have various relationships with the base product. For example, it is possible to identify existing products that are similar to the base product, as well as existing products that are dissimilar to the base product, existing products that belong to the same category as the base product, existing products that share effects with the base product, etc.

[0107] An example of identifying an existing product similar to a base product is shown in Figure 10. Specifically, in the example of Figure 10, existing product Z is identified as an existing product similar to the base product among multiple existing products shown in the existing product graph, and the existing product graph of existing product Z is extracted.

[0108] Here, the existing product graph of existing product Z extracted as described above includes a node indicating that existing product Z has the effect of "preventing drying," a node indicating that it contains an ingredient called "ingredient z," and nodes and links indicating that it is manufactured by mixing "ingredient z" for "mixing time z." These nodes show that existing product Z is similar to the base product, but has the effect of preventing drying that the base product does not have. It can be said that this effect may be due to the manufacturing method in which ingredient z, an ingredient not contained in the base product, is mixed for mixing time z.

[0109] Based on the above, in the example of Fig. 10, a manufacturing method is recommended in which ingredient z is mixed for mixing time z to add the effect of preventing dryness to the base product. In this way, the cosmetics manufacturing support method according to this exemplary embodiment makes it possible to estimate and recommend a manufacturing method for a new product that meets a user's request.

[0110] (Device configuration) The configuration of a cosmetics manufacturing support device 3 according to a third exemplary embodiment of the present invention will be described with reference to Fig. 11. Fig. 11 is a block diagram showing the configuration of a cosmetics manufacturing support device 3 according to this exemplary embodiment.

[0111] As shown in the figure, the cosmetics manufacturing support device 3 includes a reception unit 301, a graph generation unit 302, a link prediction unit 303, a property prediction unit 304, an estimation unit 305, a basis generation unit 306, and an output unit 307. Furthermore, similar to the cosmetics manufacturing support device 2 of exemplary embodiment 2, the cosmetics manufacturing support device 3 may include, in addition to these components, a learning unit, an input device, an output device, a communication device, and the like.

[0112] The receiving unit 301 receives requests for the development of new cosmetic products. This request may include, for example, the efficacy required for the new product, the raw materials or ingredients to be used in the new product, the manufacturing method to be used in manufacturing the new product, etc. This request may also include, for example, information specifying the base product that will be the basis for the new product, its ingredients, efficacy, manufacturing method, etc.

[0113] Based on the information about the base product, the graph generation unit 302 generates a base product graph that represents the base product as a graph. Specifically, the graph generation unit 302 generates a base product graph that represents the base product as nodes that represent the ingredients, effects, or manufacturing method of the base product, and edges that represent the relationships between the nodes. Note that if the base product graph is included in the existing product graph, the graph generation unit 302 can simply extract the base product graph from the existing product graph.

[0114] The link prediction unit 303 uses the base product graph and the existing product graph to identify existing products similar to the base product through link prediction, which predicts the relationship between nodes that are not connected by links in the base product graph and the existing product graph. Hereinafter, existing products similar to the base product will be referred to as "products similar to the base product." As described above, link prediction can identify existing products that have any relationship with the base product, in addition to similar relationships.

[0115] The property prediction unit 304 predicts properties that can be added to the base product. Specifically, the property prediction unit 304 refers to the existing product graph of the similar product to the base product identified by the link prediction unit 303, and predicts, among the properties shown in the existing product graph, properties that are not present in the base product as properties that can be added to the base product.

[0116] The estimation unit 305 estimates a manufacturing method for a new product that matches a request received by the reception unit 301, based on a trained model that has learned the relationship between the ingredients and effects of existing products and the manufacturing methods of the existing products, and the request. Specifically, the trained model is the above-mentioned existing product graph. As described above, the link prediction unit 303 performs link prediction using the existing product graph, and therefore the estimation unit 305 performs the above estimation based on the result of link prediction by the link prediction unit 303, thereby performing estimation based on the trained model.

[0117] The basis generating unit 306 generates basis information indicating the basis of the estimation by the estimating unit 305. The basis generating unit 306 is similar to the basis information generating unit 209 in the second exemplary embodiment, and therefore detailed description thereof will not be repeated.

[0118] The output unit 307 outputs various information generated by the cosmetics manufacturing support device 3, such as information indicating the manufacturing method estimated by the estimation unit 305. As with the output unit 210 in exemplary embodiment 2, the destination of the information output is not particularly limited.

[0119] As described above, the cosmetics manufacturing support device 3 includes a link prediction unit 303 that identifies existing products that have a predetermined relationship with the base product by link prediction, using a base product graph including multiple nodes related to the manufacturing of a base product that serves as the basis for the new product to be developed, and an existing product graph, to predict the relationship between nodes that are not connected by links in the base product graph and the existing product graph.The estimation unit 305 then estimates a manufacturing method for a new product that is based on the base product and matches the request, based on the manufacturing method for the existing product identified by the link prediction unit 303.

[0120] The manufacturing method of an existing product that has a predetermined relationship with the base product is useful information for estimating a manufacturing method of a new product that is based on the base product and that matches the request. Therefore, with the above configuration, it is possible to estimate a manufacturing method of a new product that matches the request by taking this useful information into consideration.

[0121] (Processing flow) The flow of the process (cosmetics manufacturing support method) executed by cosmetics manufacturing support device 3 will be described with reference to Fig. 12. Fig. 12 is a flowchart showing the flow of the process executed by cosmetics manufacturing support device 3.

[0122] In S301, the receiving unit 301 receives a request for the development of a new cosmetic product. Here, it is assumed that the received request includes information about a base product. The information about the base product only needs to be sufficient to generate a base product graph in the next step S302. For example, the received request may include information indicating the ingredients, efficacy, manufacturing method, etc. of the base product.

[0123] In S302, the graph generation unit 302 generates a base product graph based on the information input in S301. For example, when input of the ingredients, efficacy, and manufacturing method of a base product is received in S301, the graph generation unit 302 may generate a base product graph including nodes and links indicating the ingredients, efficacy, and manufacturing method.

[0124] In S303, the link prediction unit 303 performs link prediction using the pre-trained existing product graph and the base product graph generated in S302, and identifies products similar to the base product. Here, the basis generation unit 306 may generate basis information indicating the basis for the identification result of the link prediction unit 303 by analyzing the base product graph and the existing product graph.

[0125] In S304, the property prediction unit 304 predicts properties that can be added to the base product based on the identification result of S303. Specifically, the property prediction unit 304 refers to the existing product graph of a product similar to the base product identified in S303, and predicts, among the properties shown in the existing product graph, properties that are not present in the base product as properties that can be added to the base product.

[0126] Here, the output unit 307 may output the properties predicted by the property prediction unit 304 and present them to the user, allowing the user to select the properties desired for the new product from the presented properties. In this case, the receiving unit 301 receives the properties selected by the user as a request.

[0127] Furthermore, if the request received in S301 includes an efficacy desired for the new product, the property prediction unit 304 may determine whether the requested efficacy is included in the predicted properties after the processing of S304. If the property prediction unit 304 determines that the requested efficacy is not included, it may output to the output unit 307 a message that a manufacturing method for the new product having the requested efficacy could not be estimated, and the processing of FIG. 12 may end.

[0128] In S305, the estimation unit 305 estimates a manufacturing method for the new product that matches the request. Specifically, the estimation unit 305 first refers to the existing product graph and identifies existing products that have properties that match the request from among the existing products identified in S303 as being similar to the base product. Note that the request here refers to the request received in S301 or a request based on the prediction result in S304. In addition, the estimation unit 305 does not need to identify existing products that have all of the requested properties; it is sufficient to identify existing products that have at least some of the requested properties.

[0129] The estimation unit 305 then references the existing product graph of the identified existing product and estimates, among the elements included in the manufacturing method of the existing product, elements that are not included in the manufacturing method of the base product, as elements of the manufacturing method of the new product. The manufacturing method including these elements is the estimated result of the manufacturing method of the new product that matches the request.

[0130] In S306, the evidence generation unit 306 generates evidence information that indicates the evidence of the estimation in S306. Specifically, the evidence generation unit 306 may generate evidence information that includes information about existing products that have the same efficacy as the efficacy desired for the new product, and this evidence information may include information that indicates a manufacturing method for the existing products that have the same efficacy as the efficacy desired for the new product. For example, the evidence generation unit 306 may generate, as evidence information, the product name of the existing product identified in S305 (an existing product that has properties that match the request among existing products identified as being similar to the base product) and its manufacturing method that is shown in the existing product graph.

[0131] In S307, the output unit 307 outputs information indicating the manufacturing method estimated in S305. At this time, the output unit 307 may also output the basis information generated in S306. This completes the processing of FIG. 12.

[0132] As described above, the estimation unit 305 may estimate, as an element of the manufacturing method for the new product, an element that is not included in the manufacturing method for the base product, among the elements included in the manufacturing method for a product similar to the base product identified by the link prediction unit 303 in S303.

[0133] An element that is included in the manufacturing method of an existing product similar to the base product but is not included in the manufacturing method of the base product is likely to be an element that is compatible with the manufacturing method of the base product. Therefore, with the above configuration, an element that is likely to be compatible with the manufacturing method of the base product can be estimated as an element of the manufacturing method of the new product.

[0134] Various conditions can be set for the link prediction in S303. For example, the link prediction unit 303 may identify an existing product that is similar to the base product and has at least one of the properties required for the new product. The properties required for the new product may be identified based on the request received in S301.

[0135] Such information about existing products is extremely useful in estimating a manufacturing method for a new product having the desired properties. Therefore, this configuration allows for accurate estimation of a manufacturing method for a cosmetic product having the desired properties. Note that if S303 is configured to identify an existing product having the desired properties, the process of S304 may be omitted.

[0136] Furthermore, in S303, the link prediction unit 303 may identify a manufacturing method for an existing product that is similar to the base product and has properties that match the request by link prediction using the base product graph and the existing product graph. In this case, the estimation unit 305 may estimate elements included in the manufacturing method identified in this way as elements of the manufacturing method for the new product. Even without going through the process of identifying similar products, similar relationships between products are learned when learning the existing product graph, so it is possible to identify a manufacturing method for an existing product that has properties that match the request by the above link prediction. In this case, the processes of S303 to S305 are omitted.

[0137] Exemplary Embodiment 4 (overview) 13 is a diagram illustrating an overview of a cosmetics manufacturing support method according to this exemplary embodiment. In this exemplary embodiment, an example is described in which a manufacturing method for a new product that matches a request is estimated while a new product graph is updated.

[0138] In this exemplary embodiment, link prediction is performed using a new product graph and an existing product graph, similar to exemplary embodiment 2. The new product graph shown in the top left corner of Fig. 13 includes nodes and links indicating that the ingredients of the new product include "raw material x1" and that the mixing time of the raw materials in the manufacturing method of the new product is "mixing time x2".

[0139] Furthermore, the existing product graph shown in the upper center of Fig. 13 includes nodes and links indicating that existing product A has the effect of "preventing rough skin," that its ingredients include "ingredient a," and that the mixing time of the ingredients in its manufacturing process is "mixing time a." Similarly, the existing product graph shown in the upper right corner of Fig. 13 includes nodes and links indicating that existing product B has the effect of "preventing sunburn," that its ingredients include "ingredient b," and that the mixing time of the ingredients in its manufacturing process is "mixing time b."

[0140] By learning the existing product graphs for various existing products as described above, it becomes possible to link-predict what products are likely to have what effects. In other words, in the cosmetics manufacturing support method according to this exemplary embodiment, a tentative new product graph is generated, and link-prediction is performed on the probability that new products shown in the new product graph will have the requested effects.

[0141] For example, in the example in Figure 13, the probability that the "New Product" node in the new product graph shown on the top left will be linked to the "Prevents rough skin" node via an "Effect" link is predicted to be 30%, which is not a sufficiently high probability.

[0142] Therefore, as shown in the bottom left of the figure, the node connected to the "New Product" node in the new product graph by the "Ingredient" link is changed from "Raw Material x1" to "Raw Material x3," and link prediction is performed again. As a result, the predicted probability that the "Prevents Rough Skin" node is connected to the "New Product" node by the "Effect" link has changed to 80%.

[0143] According to the cosmetics manufacturing support method of this exemplary embodiment, the results of the above processing can be used to recommend the use of "ingredients x3" as a measure to give the new product the effect of preventing rough skin. Note that, although ingredients are replaced in the example of Figure 13, ingredients can also be added. In addition, a manufacturing method for producing a new product that meets the request can also be recommended in a similar manner.

[0144] (Device configuration) The configuration of a cosmetics manufacturing support device 4 according to a fourth exemplary embodiment of the present invention will be described with reference to Fig. 14. Fig. 14 is a block diagram showing the configuration of a cosmetics manufacturing support device 4 according to this exemplary embodiment.

[0145] As shown in the figure, the cosmetics manufacturing support device 4 includes a reception unit 401, a graph generation unit 402, a link prediction unit 403, a graph update unit 404, an estimation unit 405, a basis generation unit 406, and an output unit 407. Furthermore, similar to the cosmetics manufacturing support device 2 of exemplary embodiment 2, the cosmetics manufacturing support device 4 may include, in addition to these components, a learning unit, an input device, an output device, a communication device, and the like.

[0146] The receiving unit 401 receives a request for the development of a new cosmetic product. This request may include, for example, the efficacy desired for the new product, the raw materials or ingredients to be used in the new product, the manufacturing method to be applied when manufacturing the new product, etc. This request may also include, for example, a threshold value for the probability that the desired efficacy will be manifested in the new product.

[0147] Similar to the graph generator 204 in the second exemplary embodiment, the graph generator 402 generates a new product graph that graphically represents a new product that the user wants to manufacture, based on information about the new product.

[0148] The link prediction unit 403 uses the new product graph and the existing product graph to perform link prediction to predict the relationship between nodes that are not connected by links in the new product graph and the existing product graph, and calculates the probability that a node that exhibits characteristics that match the request will be linked to a node included in the new product graph.

[0149] The graph update unit 404 updates the new product graph. Specifically, the graph update unit 404 performs either or both of a process of replacing nodes included in the new product graph with other nodes and a process of adding new nodes to the new product graph.

[0150] The new product graph may be updated according to user input or automatically. In the former case, the graph update unit 404 may cause the output unit 407 to output a list of nodes included in the current new product graph, as well as candidate nodes to be added to the new product graph or to replace nodes in the current new product graph. In the latter case, the graph update unit 404 may select nodes to be added from these candidates. Candidate nodes may be extracted from the existing product graph of an existing product that has the requested properties.

[0151] The estimation unit 405 estimates a manufacturing method for a new product that matches a request received by the reception unit 401, based on a trained model that has learned the relationship between the ingredients and effects of existing products and the manufacturing methods of the existing products, and the request. Specifically, the trained model is the above-mentioned existing product graph. As described above, the link prediction unit 403 performs link prediction using the existing product graph, and therefore the estimation unit 405 performs the above estimation based on the result of link prediction by the link prediction unit 403, thereby performing estimation based on the trained model.

[0152] The basis generating unit 406 generates basis information indicating the basis of the estimation by the estimating unit 405. The basis generating unit 406 is similar to the basis information generating unit 209 in the second exemplary embodiment, and therefore detailed description thereof will not be repeated.

[0153] The output unit 407 outputs various information generated by the cosmetics manufacturing support device 3, such as information indicating the manufacturing method estimated by the estimation unit 405. As with the output unit 210 in exemplary embodiment 2, the destination of the information output is not particularly limited.

[0154] As described above, cosmetics manufacturing support device 4 includes link prediction unit 403 that uses a new product graph including multiple nodes related to the manufacturing of a new product to be developed and an existing product graph to calculate the probability that a node exhibiting properties that match the request will be linked to a node included in the new product graph by link prediction to predict the relationship between nodes that are not connected by links in the new product graph and the existing product graph, and estimation unit 405 estimates a manufacturing method for a new product that matches the request based on the probability calculated by link prediction unit 403.

[0155] The probability that a node indicating a property that matches the request is linked to a node included in the new product graph can be said to indicate the possibility that the new product will have the property that matches the request. Therefore, with the above configuration that estimates a manufacturing method for a new product based on this probability, it is possible to estimate a manufacturing method that will give the new product the property that matches the request.

[0156] (Processing flow) The flow of the process (cosmetics manufacturing support method) executed by cosmetics manufacturing support device 4 will be described with reference to Fig. 15. Fig. 15 is a flowchart showing the flow of the process executed by cosmetics manufacturing support device 4.

[0157] In S401, the receiving unit 401 receives a request for development of a new cosmetic product. In S401, for example, a request indicating the properties required for the new product, some of the raw materials to be used in manufacturing the new product, etc. is received.

[0158] In S402, the graph generation unit 402 generates a new product graph based on the information input in S401. For example, if an input of the efficacy required for the new product and one of the ingredients to be used in manufacturing the new product is received in S401, the graph generation unit 402 may generate a new product graph including nodes and links indicating the efficacy and ingredients.

[0159] In S403, the link prediction unit 403 calculates the probability that a node showing properties that match the request received in S401 will be linked to a node included in the new product graph generated in S402. As described above, this probability is calculated by link prediction using the learned existing product graph and the new product graph. Here, the basis generation unit 406 may generate basis information that indicates the basis for the calculation result of the link prediction unit 403 by analyzing the new product graph and the existing product graph.

[0160] In S404, the graph update unit 404 determines whether the probability calculated in S403 is equal to or greater than a threshold value. If it is determined that the probability is equal to or greater than the threshold value (YES in S404), the process proceeds to S406, and if it is determined that the probability is less than the threshold value (NO in S404), the process proceeds to S405.

[0161] If the request received in S401 specifies multiple properties, predictions are made for each property in S403, and if the probability for all properties is equal to or greater than the threshold, a YES result is made in S404, and if even one property is below the threshold, a NO result is made. This makes it possible to estimate a manufacturing method for a new product that satisfies all of the requested properties.

[0162] The request may also indicate that it does not have a predetermined property, such as "has no side effects." In this case, in S403, the probability that nodes with that property are linked may be calculated, and in S404, it may be determined whether or not the probability is equal to or less than a threshold value.

[0163] In S405, the graph update unit 404 updates the new product graph. Specifically, the graph update unit 404 may replace a node included in the current new product graph with another node, or may add a new node and link to the current new product graph. As described above, the update content may be determined according to a user input, or may be determined by the graph update unit 404.

[0164] Once the new product graph is updated, the process returns to S403, and the probability is calculated again. That is, in the process of Fig. 15, the calculation of the probability in S403 and the update of the new product graph in S405 are repeated until the determination in S404 is YES.

[0165] In S406, the estimation unit 405 estimates a manufacturing method for the new product that matches the request received in S401. Specifically, the estimation unit 405 estimates that the manufacturing method shown in the new product graph when the determination in S404 is YES is the manufacturing method for the new product that matches the request.

[0166] In S407, the evidence generating unit 406 generates evidence information indicating the evidence of the estimation in S406. Specifically, the evidence generating unit 406 may generate evidence information including information on an existing product that has the same efficacy as the efficacy required for the new product, and this evidence information may include information indicating a manufacturing method for the existing product that has the same efficacy as the efficacy required for the new product.

[0167] In S408, the output unit 407 outputs information indicating the manufacturing method estimated in S406. At this time, the output unit 407 may also output the basis information generated in S407. This ends the processing of FIG.

[0168] Exemplary Embodiment 5 (overview) 16 is a diagram illustrating an overview of a cosmetics manufacturing support method according to this exemplary embodiment. In this exemplary embodiment, an example is described in which a manufacturing method for a new product that meets a request is estimated while successively adding nodes indicating ingredients and manufacturing methods to a new product graph.

[0169] In this exemplary embodiment, link prediction is performed using a new product graph and an existing product graph, similar to exemplary embodiment 2. The new product graph shown in the top left corner of Fig. 16 includes nodes and links indicating that the ingredients of the new product include "ingredient x1" and that the new product is required to have the effect of "preventing rough skin."

[0170] Furthermore, the existing product graph shown in the upper center of Fig. 16 includes nodes and links indicating that existing product A has the effect of "preventing rough skin," that its ingredients include "ingredient a," and that the mixing time of the ingredients in its manufacturing process is "mixing time a." Similarly, the existing product graph shown in the upper right corner of Fig. 13 includes nodes and links indicating that existing product B has the effect of "preventing sunburn," that its ingredients include "ingredient b," and that the mixing time of the ingredients in its manufacturing process is "mixing time b."

[0171] By learning the existing product graphs for various existing products as described above, it becomes possible to perform link prediction to determine which ingredients and manufacturing methods are likely to be applicable to which products. In other words, the cosmetics manufacturing support method according to this exemplary embodiment predicts ingredients and manufacturing methods that are likely to be applicable to a new product through link prediction, and estimates the manufacturing method for the new product by adding nodes to the new product graph based on the prediction results.

[0172] For example, in the example of Figure 16, a link prediction is made for the node connected by the "ingredient" link to the "new product" node in the new product graph shown at the top left. Then, as shown at the bottom left, this link prediction predicts that the node connected by the "ingredient" link to the "new product" node is "ingredient a," and based on this, this node is added to the new product graph.

[0173] After determining the ingredients of the new product in this way, link prediction is performed on the nodes connected to the "new product" node by the "manufacturing method" link in the new product graph after the nodes have been added. This makes it possible to predict the manufacturing method taking into account the nodes indicating the ingredients that have been added based on the results of link prediction.

[0174] According to the cosmetics manufacturing support method of this exemplary embodiment, by repeating the above process, it is possible to sequentially predict the ingredients and manufacturing methods that will give a new product the effect of preventing rough skin, and to recommend manufacturing methods that include each of the predicted ingredients and manufacturing methods.

[0175] (Device configuration) The configuration of the cosmetics manufacturing support device according to the fifth exemplary embodiment of the present invention is the same as the configuration of the cosmetics manufacturing support device 4 shown in Fig. 14. The difference lies in the processing content executed by the link prediction unit 403 and the estimation unit 405.

[0176] Specifically, in the cosmetics manufacturing support device according to the fifth exemplary embodiment, the link prediction unit 403 uses a new product graph and an existing product graph, each of which includes multiple nodes related to the manufacturing of a new product to be developed, to predict the relationships between nodes not connected by links in the new product graph and the existing product graph, and predicts nodes that link to nodes in the new product graph from among the nodes in the existing product graph that indicate manufacturing methods for existing products.The estimation unit 405 then estimates a manufacturing method for the new product that matches the request based on the nodes predicted by the link prediction unit.

[0177] Among the nodes that indicate manufacturing methods for existing products, the nodes that link to the new product graph can be said to indicate manufacturing methods that are likely to be compatible with the manufacturing method for the new product. Therefore, with the above configuration that estimates the manufacturing method for the new product based on these nodes, it is possible to estimate the manufacturing method that is likely to be compatible with the new product.

[0178] (Processing flow) The flow of processing (cosmetics manufacturing support method) executed by a cosmetics manufacturing support device according to a fifth exemplary embodiment of the present invention will be described with reference to Figure 17. Figure 17 is a flow diagram showing the flow of processing executed by a cosmetics manufacturing support device according to a fifth exemplary embodiment of the present invention. Note that, as in the fourth exemplary embodiment, the processing will be executed by receiving unit 401 to output unit 407 (see Figure 14).

[0179] In S501, the receiving unit 401 receives a request for development of a new cosmetic product. In S401, for example, a request indicating the properties required for the new product, some of the raw materials to be used in manufacturing the new product, etc. is received.

[0180] In S502, the graph generation unit 402 generates a new product graph based on the information input in S501. For example, if an input of the efficacy required for the new product and one of the ingredients to be used in manufacturing the new product is received in S401, the graph generation unit 402 may generate a new product graph including nodes and links indicating the efficacy and ingredients.

[0181] In S502, a manufacturing method for the new product can be estimated by generating a new product graph for the new product that satisfies all of the properties indicated in the request received in S501. Note that the new product graph at this stage only needs to include nodes and links that indicate that the new product satisfies the above properties.

[0182] In S503, the link prediction unit 403 predicts nodes that link to nodes included in the new product graph generated in S502. As described above, this node is predicted by link prediction using the learned existing product graph and the new product graph. Here, the basis generation unit 406 may generate basis information that indicates the basis for the prediction result of the link prediction unit 403 by analyzing the new product graph and the existing product graph.

[0183] In S503, the link prediction unit 403 may predict a node connected to the "new product" node via a link of "ingredients" or "manufacturing method." Alternatively, it may predict a node connected to a node indicating ingredients or a node indicating a manufacturing method. For example, it may predict a node connected to a node indicating ingredients via a link of "mixing amount," or it may predict a node connected to a node indicating mixing time via a link of "mixing speed." This makes it possible to estimate details of the ingredients and manufacturing method of the new product.

[0184] In S504, the graph update unit 404 determines whether or not to confirm the manufacturing method for the new product. If it is determined that the manufacturing method is confirmed (YES in S504), the process proceeds to S506, and if it is determined that the manufacturing method is not confirmed (NO in S504), the process proceeds to S505.

[0185] The conditions for determining the manufacturing method in S504 may be determined in advance. For example, thresholds may be set for the number of nodes indicating ingredients and the number of nodes indicating manufacturing methods, and the manufacturing method may be determined when the number of these nodes included in the new product graph both exceeds the threshold. Furthermore, for example, whether or not to determine the manufacturing method may be determined according to a user input. In this case, it is preferable that the output unit 407 outputs the new product graph or the ingredients and manufacturing methods shown therein, and presents this information to the user.

[0186] In S505, the graph update unit 404 adds nodes and links to the new product graph. The links and nodes to be added may be determined according to user input, or may be determined by the graph update unit 404. Note that in S505, the graph update unit 404 may replace a node included in the current new product graph with another node according to user input.

[0187] Once the new product graph has been updated with the addition of nodes and links, the process returns to S503, where nodes that link to nodes included in the updated new product graph are predicted. That is, in the process of Figure 17, node prediction in S503 and updating of the new product graph are repeated until a YES determination is made in S504. The processes of S506 to S508 are the same as the processes of S406 to S408 in Figure 15.

[0188] [Modification] As described above, by using a new product graph and an existing product graph, it is possible to predict the properties of a new product shown in the new product graph by link prediction. Furthermore, new product property prediction can also be performed by methods other than link prediction. This will be explained with reference to FIG. 18. FIG. 18 is a diagram illustrating an example of predicting the properties of a new product based on feature quantities calculated from the new product graph and the existing product graph. FIG. 18 shows existing product graphs for existing products A to C and a new product graph for a new product. Note that the nodes and links included in these graphs are omitted from the illustration.

[0189] Here, the feature quantities for each existing product can be calculated by multiplying the feature quantities of each node included in the existing product graph by a weight corresponding to the link connected to that node and adding the results together. Therefore, if learning is performed to update the weights so that the calculated feature quantities correspond to the properties of the existing products, it becomes possible to predict the properties of new products from the feature quantities of the new product graph calculated by applying the weights.

[0190] For example, in the example of Fig. 18, the feature amount calculated from the existing product graph of existing product A, which is known to be gentle on the skin, is trained to fall within a range in the feature space corresponding to the property "gentle on the skin." Also, the feature amount calculated from the existing product graph of existing product B, which is known to have the effect of preventing rough skin, is trained to fall within a range in the feature space corresponding to the property "prevents rough skin."

[0191] In this case, as shown in the figure, if the feature quantity calculated from the new product graph is within the range corresponding to the attribute "gentle on skin," it can be predicted that the new product has the attribute "gentle on skin." Such an attribute prediction method can be applied as an alternative method to the attribute prediction methods in the above-described exemplary embodiments.

[0192] [Software implementation example] Some or all of the functions of the cosmetics manufacturing support devices 1 to 4 and the cosmetics manufacturing support device of exemplary embodiment 5 (hereinafter referred to as the devices) may be realized by hardware such as an integrated circuit (IC chip), or by software.

[0193] In the latter case, the device is realized, for example, by a computer that executes instructions in a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 19. Computer C includes at least one processor C1 and at least one memory C2. Memory C2 stores a program (cosmetics manufacturing support program) P for operating computer C as the device. In computer C, processor C1 reads and executes program P from memory C2, thereby realizing each function of the device.

[0194] The processor C1 may be, for example, a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0195] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may also include a communication interface for transmitting and receiving data to and from other devices. The computer C may also include an input / output interface for connecting input / output devices such as a keyboard, mouse, display, and printer.

[0196] Furthermore, the program P can be recorded on a non-transitory tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can acquire the program P via such a recording medium M. The program P can also be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also acquire the program P via such a transmission medium.

[0197] [Appendix 1] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means disclosed in the above-described embodiments are also included in the technical scope of the present invention.

[0198] [Appendix 2] Some or all of the above-described embodiments can also be described as follows: However, the present invention is not limited to the following described aspects.

[0199] (Appendix 1) A cosmetics manufacturing support device comprising: a receiving means for receiving a request for development of a new cosmetic product; a trained model that has learned the relationship between the ingredients and effects of existing cosmetic products and manufacturing methods of the existing products; an estimation means for estimating a manufacturing method of a new product that matches the request based on the request; and an output means for outputting information indicating the manufacturing method estimated by the estimation means. This configuration provides the effect of being able to effectively support the development of new cosmetic products.

[0200] (Appendix 2) The cosmetics manufacturing support device according to claim 1, wherein the request includes information indicating the efficacy desired for the new product, and the device further includes a basis generating means for generating, as the basis for the estimation, evidence information including information about the existing product that has the same efficacy as the efficacy desired for the new product, and the output means further outputs the evidence information. This configuration allows the user to be provided with valuable information for determining whether the estimated manufacturing method is effective for imparting the desired efficacy to the new product.

[0201] (Appendix 3) The cosmetics manufacturing support device according to claim 2, wherein the evidence information includes information indicating a manufacturing method for the existing product that has the same efficacy as the efficacy required for the new product. This configuration provides a user with valuable information to determine whether the estimated manufacturing method is effective for imparting the desired efficacy to the new product.

[0202] (Appendix 4) The cosmetics manufacturing support device according to Supplementary Note 1, wherein the trained model is an existing product graph in which one or more of the existing products are represented by nodes representing the ingredients, efficacy, or manufacturing methods of the existing products, and edges representing the relationships between the nodes. With this configuration, it is possible to estimate an appropriate manufacturing method by taking into account the mutual relationships between the ingredients, efficacy, and manufacturing methods of the existing products.

[0203] (Appendix 5) The cosmetics manufacturing support device according to claim 4, wherein the request includes information indicating the efficacy desired for the new product, and the estimation means estimates, as a manufacturing method for manufacturing the new product having the desired efficacy, a manufacturing method that includes some elements of a manufacturing method for an existing product that corresponds to a node and link that indicate the efficacy included in the existing product graph. With this configuration, it is possible to estimate a manufacturing method for a new cosmetic product having the desired efficacy.

[0204] (Appendix 6) The cosmetics manufacturing support device according to Supplementary Note 4 includes a link prediction means for identifying existing products having a predetermined relationship with the new product by link prediction using a new product graph including a plurality of nodes related to the manufacturing of a new product to be developed and the existing product graph to predict relationships between nodes not connected by links in the new product graph and the existing product graph, and the estimation means for estimating a manufacturing method for the new product that matches the request based on the manufacturing method for the existing product identified by the link prediction means. With this configuration, it is possible to estimate a manufacturing method for the new product that matches the request by taking this useful information into consideration.

[0205] (Appendix 7) The cosmetics manufacturing support device according to Supplementary Note 6 further comprises an existing product identification means for identifying the existing product having properties that match the request based on the nodes and links included in the existing product graph. With this configuration, it is possible to estimate a manufacturing method for a new product that matches the request by taking this useful information into consideration.

[0206] (Appendix 8) The cosmetics manufacturing support device according to claim 6 or 7, wherein the link prediction means identifies existing products similar to the new product by the link prediction, and includes a property prediction means that predicts the properties of the existing products identified by the link prediction means and shown in the existing product graph as the properties of the new product. This configuration makes it possible to predict appropriate properties of the new product.

[0207] (Appendix 9) The cosmetics manufacturing support device according to Supplementary Note 4 includes a link prediction means for identifying existing products that have a predetermined relationship with the base product by link prediction using a base product graph including multiple nodes related to the manufacture of a base product on which the new product to be developed is based and the existing product graph to predict relationships between nodes that are not connected by links in the base product graph and the existing product graph, and the estimation means for estimating a manufacturing method for a new product that is based on the base product and that matches the request, based on the manufacturing method for the existing product identified by the link prediction means. With this configuration, it is possible to estimate a manufacturing method for a new product that matches the request, taking into account this useful information.

[0208] (Appendix 10) The cosmetics manufacturing support device according to claim 9, wherein the link prediction means identifies the existing product that is similar to the base product and has at least one property required for the new product. This configuration makes it possible to accurately estimate a manufacturing method for cosmetics having desired properties.

[0209] (Appendix 11) The cosmetics manufacturing support device according to claim 9 or 10, wherein the estimation means estimates, as elements of the manufacturing method for the new product, elements that are not included in the manufacturing method for the base product, among elements included in the manufacturing method for the existing product identified by the link prediction means. With this configuration, elements that are likely to be compatible with the manufacturing method for the base product can be estimated as elements of the manufacturing method for the new product.

[0210] (Appendix 12) The cosmetics manufacturing support device according to Supplementary Note 4 includes a link prediction means that calculates the probability that a node showing properties that match the request will be linked to a node included in the new product graph by link prediction using a new product graph including multiple nodes related to the manufacturing of a new product to be developed and the existing product graph to predict relationships between nodes that are not connected by links in the new product graph and the existing product graph, and the estimation means estimates a manufacturing method for the new product that matches the request based on the probability calculated by the link prediction means. With this configuration, it is possible to estimate a manufacturing method that will impart properties that match the request to the new product.

[0211] (Appendix 13) The cosmetics manufacturing support device according to claim 4 further comprises a link prediction means for predicting, by link prediction using the trained model including a new product graph including a plurality of nodes related to the manufacturing of a new product to be developed and the existing product graph, relationships between nodes not connected by links in the new product graph and the existing product graph, from among nodes included in the existing product graph that indicate manufacturing methods for the existing products, nodes that will link to the new product graph, and the estimation means for estimating a manufacturing method for a new product that matches the request based on the nodes predicted by the link prediction means. This configuration makes it possible to estimate a manufacturing method that is likely to be suitable for a new product.

[0212] (Appendix 14) A cosmetics manufacturing support method in which a computer receives a request for the development of a new cosmetic product, estimates a manufacturing method for the new product that matches the request based on the request and a trained model that has learned the relationship between the ingredients and efficacy of existing cosmetic products and the manufacturing methods of the existing products, and outputs information indicating the estimated manufacturing method. This configuration has the effect of being able to provide optimal support for the development of new cosmetic products.

[0213] (Appendix 15) A cosmetics manufacturing support program that causes a computer to execute the following processes: accepting a request for the development of a new cosmetic product; estimating a manufacturing method for a new product that matches the request based on a trained model that has learned the relationship between the ingredients and efficacy of existing cosmetic products and manufacturing methods of the existing products; and outputting information indicating the estimated manufacturing method. This configuration provides the effect of being able to effectively support the development of new cosmetic products.

[0214] [Appendix 3] Some or all of the above-described embodiments can also be expressed as follows.

[0215] A cosmetics manufacturing support device comprising at least one processor that executes the following processes: accepting a request for the development of a new cosmetic product; estimating a manufacturing method for a new product that matches the request based on a trained model that has learned the relationship between the ingredients and efficacy of an existing cosmetic product, which is an existing cosmetic product, and the manufacturing method of the existing product, and the request; and outputting information indicating the estimated manufacturing method.

[0216] The cosmetics manufacturing support device may further include a memory that stores a program (cosmetics manufacturing support program) that causes the processor to execute the following steps: accepting a request for development of a new cosmetic product; estimating a manufacturing method for a new product that matches the request based on a trained model that has learned the relationship between the ingredients and effects of existing cosmetic products and manufacturing methods for the existing products; and outputting information indicating the estimated manufacturing method. The program may also be recorded on a computer-readable, non-transitory, tangible recording medium. [Explanation of symbols]

[0217] 1. Cosmetic manufacturing support equipment 11 Reception 12 Estimation part 13 Output section 2. Cosmetic manufacturing support equipment 201 Reception 202 Existing Product Identification Department 203 Evaluation Department 205 Link Prediction Unit 207 Property Prediction Department 208 Estimation Department 209 Evidence Generation Unit 210 Output section 3 Cosmetic manufacturing support equipment 301 Reception 303 Link Prediction Unit 304 Property Prediction Department 305 Estimation Department 306 Evidence Generation Unit 307 Output section 4. Cosmetic manufacturing support equipment 401 Reception 403 Link Prediction Unit 405 Estimation section 406 Evidence Generation Unit 407 Output Section

Claims

1. A means for receiving requests regarding the development of new cosmetic products; an estimation means for estimating a manufacturing method of a new product that conforms to the request based on a trained model that has learned the relationship between the ingredients and efficacy of an existing product, which is an existing cosmetic product, and a manufacturing method of the existing product, and the request; an output means for outputting information indicating the manufacturing method estimated by the estimation means; Equipped with The manufacturing method includes a processing method indicating at least one of a pre-processing method of raw materials, a mixing time, a mixing speed, and a post-processing method, and at least one of equipment or a device used in the manufacturing method. Cosmetic manufacturing support equipment.

2. the request includes information indicating a desired efficacy of the new product; a basis generating means for generating, as a basis for the estimation, basis information including information on the existing product having the same efficacy as the efficacy required for the new product; The output means further outputs the basis information. The cosmetics manufacturing support device according to claim 1 .

3. The evidence information includes information indicating a manufacturing method of the existing product having the same efficacy as the efficacy required for the new product. The cosmetics manufacturing support device according to claim 2.

4. The trained model is an existing product graph in which one or more of the existing products are represented by nodes representing the ingredients, efficacy, or manufacturing method of the existing product, and edges representing the relationships between the nodes. The cosmetics manufacturing support device according to claim 1 .

5. the request includes information indicating a desired benefit of the new product; 5. The cosmetics manufacturing support device according to claim 4, wherein the estimation means estimates, as a manufacturing method for manufacturing the new product having the efficacy, a manufacturing method that includes some elements of a manufacturing method for an existing product that corresponds to a node and link that indicate the efficacy included in the existing product graph.

6. a link prediction means for using a new product graph including a plurality of nodes related to the manufacture of a new product to be developed and the existing product graph to predict a relationship between nodes that are not connected by a link in the new product graph and the existing product graph, thereby identifying the existing products that have a predetermined relationship with the new product; The cosmetics manufacturing support device according to claim 4 , wherein the estimation means estimates a manufacturing method for the new product that matches the request based on the manufacturing method for the existing product identified by the link prediction means.

7. The cosmetics manufacturing support device according to claim 6 , further comprising an existing product identification means for identifying the existing product having properties that match the request based on the nodes and links included in the existing product graph.

8. the link prediction means identifies the existing product similar to the new product through the link prediction; 8. The cosmetics manufacturing support device according to claim 6, further comprising a property prediction means for predicting the properties of the existing products identified by the link prediction means and shown in the existing product graph as the properties of the new product.

9. a link prediction means for using a base product graph including a plurality of nodes related to the manufacture of a base product on which a new product to be developed is based and the existing product graph to predict a relationship between nodes not connected by a link in the base product graph and the existing product graph, thereby identifying the existing products having a predetermined relationship with the base product; The estimation means estimates a manufacturing method of a new product that matches the request, based on the base product, based on the manufacturing method of the existing product identified by the link prediction means. The cosmetics manufacturing support device according to claim 4.

10. The cosmetics manufacturing support device according to claim 9 , wherein the link prediction means identifies the existing product that is similar to the base product and has at least one property required for the new product.

11. 11. The cosmetics manufacturing support device according to claim 9, wherein the estimation means estimates, among elements included in the manufacturing method of the existing product identified by the link prediction means, elements that are not included in the manufacturing method of the base product as elements of the manufacturing method of the new product.

12. a link prediction means for calculating the probability that a node showing a property that matches the request will be linked to a node included in the new product graph by link prediction for predicting a relationship between nodes that are not connected by a link in the new product graph and the existing product graph, using the new product graph including a plurality of nodes related to the manufacture of the new product to be developed and the existing product graph; The estimation means estimates a manufacturing method of a new product that matches the request based on the probability calculated by the link prediction means. The cosmetics manufacturing support device according to claim 4.

13. a link prediction means for predicting, by link prediction for predicting relationships between nodes not connected by links in the new product graph and the existing product graph using the learned model including a new product graph including a plurality of nodes related to the manufacturing of a new product to be developed and the existing product graph, nodes that are linked to the new product graph from among nodes included in the existing product graph that indicate manufacturing methods of the existing products; The estimation means estimates a manufacturing method of a new product that matches the request based on the nodes predicted by the link prediction means. The cosmetics manufacturing support device according to claim 4.

14. The computer We accept requests for new cosmetic product development, A manufacturing method of a new product that conforms to the request is estimated based on the request and a trained model that has learned the relationship between the ingredients and efficacy of an existing product, which is an existing cosmetic product, and the manufacturing method of the existing product; outputting information indicating the estimated manufacturing method; The manufacturing method includes a processing method indicating at least one of a pre-processing method of raw materials, a mixing time, a mixing speed, and a post-processing method, and at least one of equipment or a device used in the manufacturing method. Cosmetic manufacturing support methods.

15. For computers, Processing of receiving requests for new cosmetic product development; A process of estimating a manufacturing method of a new product that conforms to the request based on a trained model that has learned the relationship between the ingredients and efficacy of an existing product, which is an existing cosmetic product, and the manufacturing method of the existing product, and the request; a process of outputting information indicating the estimated manufacturing method; Execute The manufacturing method includes a processing method indicating at least one of a pre-processing method of raw materials, a mixing time, a mixing speed, and a post-processing method, and at least one of equipment or a device used in the manufacturing method. Cosmetics manufacturing support program.

Images