Search system, search method, and search program

Abstract

This search system: acquires a composition target value designated by a user; acquires, for each of a plurality of compositions, composition data indicating one or more raw materials and one or more characteristic values of the composition; acquires, for each of the plurality of raw materials, raw material data indicating one or more attribute values of the raw material; executes association analysis based on the target value, the composition data, and the raw material data to identify, as a first raw material group, one or more raw materials which correspond to the target value; inputs each of one or more data records of the raw material data to a trained model to calculate, for each of the one or more data records, a degree of correspondence between the raw material indicated by the data record and the target value; and on the basis of the one or more degrees of correspondence, identifies, as a second raw material group, one or more raw materials which correspond to the target value and which do not belong to the first raw material group.

Description

Search system, search method, and search program

[0001] One aspect of the present disclosure relates to a search system, a search method, and a search program.

[0002] A computer system that provides reference information for producing a composition is known. For example, Patent Document 1 describes a production evaluation system that includes a production device that produces a sample and a sample information management device that manages information about the sample produced by the production device. The sample information management device includes a sample information storage unit that stores sample information about the sample produced by the production device, a recommendation target determination unit that references the sample information and determines, as a target for recommendation, production conditions corresponding to material information that meets predetermined criteria, and a recommendation unit that provides the determined production conditions to a user.

[0003] JP 2023-86450 A

[0004] There is a need for a system for comprehensively and efficiently searching for raw materials that are expected to produce a desired composition.

[0005] A search system according to one aspect of the present disclosure includes at least one processor, which acquires a target value for a composition specified by a user, acquires, for each of a plurality of compositions, composition data indicating one or more ingredients and one or more characteristic values ​​of the composition, acquires, for each of a plurality of ingredients, ingredient data indicating one or more attribute values ​​of the ingredient, performs association analysis based on the target value, the composition data, and the ingredient data to identify one or more ingredients corresponding to the target value as a first ingredient group, inputs each of one or more data records of the ingredient data into a trained model that is trained to accept input of data records of the ingredient data and output the degree of correspondence between the ingredients indicated in the data records and the target value, calculates one or more degrees of correspondence corresponding to the one or more data records, and identifies, based on the one or more degrees of correspondence, one or more ingredients that correspond to the target value and do not belong to the first ingredient group as a second ingredient group, and presents the first ingredient group and the second ingredient group to the user.

[0006] In this respect, ingredients that are expected to achieve the target composition values ​​are identified using two techniques: association analysis and trained models, so that such ingredients can be presented to the user comprehensively and efficiently.

[0007] According to one aspect of the present disclosure, it is possible to comprehensively and efficiently search for raw materials that are expected to be able to realize a desired composition.

[0008] FIG. 1 is a diagram showing an example of the functional configuration of a search system; FIG. 2 is a diagram showing an example of a raw material database; FIG. 3 is a diagram showing an example of a composition database; FIG. 4 is a flowchart showing an example of processing by the search system; FIG. 5 is a flowchart showing an example of a first search; FIG. 6 is a flowchart showing an example of a second search; and FIG. 7 is a diagram showing an example of a screen for using the search system.

[0009] Various examples of the present disclosure will be described in detail below with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are designated by the same reference numerals, and redundant description will be omitted.

[0010] [System Configuration] The search system according to the present disclosure is a computer system that searches for ingredients that are expected to produce a desired composition. The search system performs its search using two techniques: association analysis and a trained model. This mechanism enables a more comprehensive and efficient search for ingredients than would be possible using a single search technique, and the search results can be presented to the user.

[0011] Association analysis is a technique for finding relationships between data items in a large amount of data. For example, association analysis can analyze the co-occurrence of specific data values.

[0012] Trained models are generated using machine learning, a method of autonomously discovering laws or rules by repeatedly learning based on given information. Trained models are constructed using algorithms and data structures.

[0013] A search system is composed of one or more computers. When multiple computers are used, these computers are connected via a communication network such as the Internet or an intranet to logically construct a single search system.

[0014] A computer that constitutes a search system generally comprises a processor, memory, and a communication interface as hardware devices. The processor is, for example, a CPU, and the memory is composed of a flash memory, a hard disk, etc. Each function of the search system is realized by the processor executing a program stored in the memory. The computer may further comprise input devices such as a keyboard and a mouse, and output devices such as a monitor and speakers.

[0015] A search program for causing a computer to function as a search system includes program code for implementing each functional module of the search system. This search program may be provided by being non-temporarily recorded on a tangible recording medium, such as a CD-ROM, DVD-ROM, or semiconductor memory. Alternatively, the search program may be provided via a communications network as a data signal superimposed on a carrier wave. The provided search program is recorded in memory, for example.

[0016] 1 is a diagram showing the functional configuration of an example search system 10. In this example, the search system 10 is connected to an ingredient database 21, a composition database 22, and a user terminal 30 via a communication network such as the Internet or an intranet.

[0017] The raw material database 21 contains raw material data D Ｒ The composition database 22 stores composition data D Ｃ Both the raw material database 21 and the composition database 22 may be provided in a computer system different from the search system 10, or may be components of the search system 10.

[0018] The user terminal 30 is a computer used by a user of the search system 10. The user terminal 30 may be any of various computers, such as a personal computer, a workstation, a tablet terminal, a smartphone, or a wearable terminal.

[0019] The search system 10 includes a processor 101. In one example, the processor 101 functions as an acquisition unit 11, a first search unit 12, a model generation unit 13, and a second search unit 14. The acquisition unit 11 is a functional module that acquires search conditions. The first search unit 12 is a functional module that searches for ingredients that correspond to the search conditions through association analysis. The model generation unit 13 is a functional module that generates a trained model 19 used to search for ingredients that correspond to the search conditions. The second search unit 14 is a functional module that searches for ingredients that correspond to the search conditions using the trained model 19. In the present disclosure, the search by the first search unit 12 is also referred to as a "first search," and the search by the second search unit 14 is also referred to as a "second search."

[0020] FIG. 2 is a diagram showing an example of the raw material database 21. Ｒ Each data record in includes an ingredient identifier, which is an identifier for uniquely identifying each ingredient, and one or more attribute values ​​corresponding to one or more attributes of the ingredient. The attribute of an ingredient refers to the characteristics or properties of the ingredient. Examples of attributes include structural features, physical property values, and catalog data. Examples of structural features include characteristics related to double bonds, functional groups, etc. Examples of physical property values ​​include specific gravity, thermal conductivity, etc. Examples of catalog data include viscosity, effect time, etc. Attribute values ​​can be expressed in various data formats such as numerical values, text, and true / false values. If an ingredient does not correspond to some attributes, ingredient data D Ｒ In at least one data record, some attribute values ​​may be missing (null). In the present disclosure, such missing values ​​are also represented as "N / A."

[0021] FIG. 3 is a diagram showing an example of the composition database 22. ＣEach data record in includes a composition identifier, which is an identifier for uniquely identifying an individual composition, one or more ingredients of the composition, and one or more characteristic values. The ingredients and characteristic values ​​of a composition can also be said to be attributes (characteristics or properties) of the composition. The value indicating an ingredient is expressed by the blend amount, which is the amount of ingredient blended for the composition. The blend amount may be expressed as an absolute value or a relative value such as a ratio. A value greater than 0 for a certain ingredient corresponding to a certain composition indicates that the ingredient was used in producing the composition, and a value of 0 indicates that the ingredient was not used in producing the composition. Data items indicating characteristic values ​​can be expressed in various data formats such as numerical values, text, and true / false values. If a composition does not correspond to some characteristic values, for example, composition data D Ｃ In at least one data record, some characteristic values ​​may be missing values ​​"N / A."

[0022] [System Operation] An example of processing by the search system 10 will be described with reference to Figures 4 to 6, as well as an example of a search method according to the present disclosure. Figure 4 is a flowchart showing this example as processing flow S1. Figure 5 is a flowchart showing an example of a first search. Figure 6 is a flowchart showing an example of a second search.

[0023] In step S11, the acquisition unit 11 acquires the target values ​​of the composition specified by the user as search conditions. The user operates the user terminal 30 to input the target values ​​of the composition. The user may specify a target value for each of two or more data items. Each target value may be represented by a single value, a set of two or more values, or a range defined by at least one of a lower limit and an upper limit. In response to the user's operation, the user terminal 30 transmits the set of target values ​​for one or more data items as search conditions to the search system 10. The acquisition unit 11 receives the search conditions.

[0024] In step S12, the first search unit 12 executes a first search using association analysis. An example of the first search will be described in detail with reference to FIG.

[0025] In step S121, the first search unit 12 searches the composition data D Ｃ Based on the above, a composition vector T is generated, which is a vector indicating whether each of the multiple compositions satisfies the target value (search condition). Ｃ If T consists of n data records corresponding to n compositions, the dimension of the composition vector T is n. Each component of the composition vector T represents a binary value indicating whether the composition satisfies a target value. For example, the composition vector T can be represented as {TRUE, FALSE, TRUE, ..., FALSE}.

[0026] In step S122, the first search unit 12 searches the composition data D Ｃ Based on this, a composition matrix X, which is a matrix showing the relationship between the composition and the raw materials, is obtained. ｂｏｏｌ Generate the composition matrix X ｂｏｏｌ indicates, for each of a plurality of compositions, the raw materials among the plurality of raw materials that are used for the composition and the raw materials among the plurality of raw materials that are not used for the composition. That is, the composition matrix X ｂｏｏｌ indicates, for each of a plurality of compositions, the ingredients whose blending amount in the composition is not 0 and the ingredients whose blending amount in the composition is 0. ｂｏｏｌ Each row of the composition data D indicates a composition, and each column indicates a raw material. Ｃ indicates n compositions, and raw material data D Ｒ If m represents the ingredients, then the composition matrix X ｂｏｏｌ is a matrix with n rows and m columns. For example, the composition matrix X ｂｏｏｌ As shown below, TRUE indicates that the blending amount is greater than 0, and FALSE indicates that the blending amount is 0.

[0027] In step S123, the first search unit 12 searches the composition vector T and the composition matrix X ｂｏｏｌ The association analysis is performed based on the above to calculate the index of each of the ingredients. This association analysis is performed based on the event E that the composition satisfies the target value. ａ and the composition contains raw material R ｉ Event E that (i = 1 to m) is used ｂFor example, the first search unit 12 may calculate at least one of a support level and a lift value as an index. The support level indicates the rate at which two events occur simultaneously. The lift value indicates how much one event promotes the other event. The first search unit 12 searches for an event E ａ , E ｂ The support indicates the rate at which events E occur simultaneously. ｂ is event E ａ and a lift value indicating how much the movement is promoted.

[0028] In step S124, the first search unit 12 identifies one or more ingredients whose indexes satisfy a predetermined criterion as a first group of ingredients. For example, the first search unit 12 identifies one or more ingredients whose support level is equal to or exceeds a threshold value S Ｌ Alternatively, the first search unit 12 may extract raw materials whose lift value is equal to or greater than the threshold value L Ｌ Alternatively, the first search unit 12 may extract ingredients whose support is equal to or greater than a threshold value S Ｌ The lift value is equal to or greater than the threshold value L Ｌ In any case, the first search unit 12 stores the extracted data records of one or more ingredients in the ingredient data D. Ｒ Then, the first search unit 12 identifies the one or more identified raw materials as a first raw material group.

[0029] As shown in steps S121 to S124, the first search unit 12 searches for the target value, the composition data D Ｃ , and raw material data D Ｒ An association analysis based on the above is performed to identify one or more ingredients that correspond to the target value as a first ingredient group. The ingredients that correspond to the target value can be said to be ingredients that contribute to the composition satisfying the target value.

[0030] In step S125, the first search unit 12 searches the raw material data D Ｒ , and identify attribute values ​​common to the first group of ingredients as common features. The common features may be identified in one or more of the multiple attributes of the ingredients. For example, the attribute value as the common feature may be represented by a single value, a set of two or more values, or a range defined by at least one of a lower limit and an upper limit.

[0031] In step S126, the first search unit 12 presents the first search results indicating the first group of ingredients and their common characteristics to the user. The first search unit 12 transmits the first search results to the user terminal 30. The user terminal 30 receives and displays the first search results. As a result, the user can confirm the ingredients found using association analysis.

[0032] 4, in step S13, the second search unit 14 executes a second search using the trained model 19. An example of the second search will be described in detail with reference to FIG.

[0033] In step S131, the model generation unit 13 generates a vector B, which is a vector indicating whether each of the multiple ingredients is included in the first ingredient group. ｍ If the vector B is composed of m data records corresponding to m ingredients, the dimension of the vector B is m. Each component of the vector B represents a binary value indicating whether the ingredient is included in the first ingredient group. For example, the vector B can be expressed as {FALSE, FALSE, TRUE, ..., FALSE}.

[0034] In step S132, the model generating unit 13 generates the raw material data D Ｒ Among the multiple raw materials indicated by the formula (1), a set J of one or more raw materials used in at least one of the multiple compositions is identified as a raw material group used. Each raw material constituting the set J (raw material group used) is represented by the composition data D Ｃ The ingredient has a blend amount greater than 0 in at least one data record.

[0035] In step S133, the model generation unit 13 generates the raw material data D limited to the set J (a group of raw materials used). Ｒ Limited ingredient data D Ｒ（ｊ∈Ｊ） The vector B limited to the set J (raw material group) is generated as the limited raw material vector B （ｊ∈Ｊ） Generated as the limited material vector B （ｊ∈Ｊ） indicates whether each ingredient constituting the used ingredient group is included in the first ingredient group.

[0036] In step S134, the model generation unit 13 generates the limited ingredient data D Ｒ（ｊ∈Ｊ）is used as an explanatory variable, and the limited material vector B （ｊ∈Ｊ） The trained model 19 is generated by performing machine learning using the raw material data D Ｒ The classification model is trained to receive input of a data record and output the degree of correspondence between the ingredient indicated in the data record and the target value. The degree of correspondence is an index showing the degree to which the ingredient contributes to the target value. In one example, the degree of correspondence is expressed by the probability of contributing to the target value. Ｒ（ｊ∈Ｊ） may include missing values. The model generation unit 13 may generate the trained model 19 by machine learning that can directly handle missing values, such as random forest, XGBoost, etc. Alternatively, the model generation unit 13 may complement the missing values ​​by imputation such as multiple imputation, regression imputation, or representative value imputation as preprocessing, and then generate the trained model 19 by machine learning such as a neural network or logistic regression.

[0037] In step S135, the second search unit 14 searches the raw material data D Ｒ For each of the one or more data records, the data record is input to the trained model 19 to calculate the degree of correspondence of the data record. Through this process, the second search unit 14 calculates one or more degrees of correspondence corresponding to the one or more data records.

[0038] In step S136, the second search unit 14 identifies one or more ingredients that correspond to the target value and do not belong to the first ingredient group as a second ingredient group based on a correspondence degree of one or more. The second ingredient group can also be said to be a collection of ingredients that have similar characteristics to the first ingredient group and are expected to contribute to the target value. In one example, the second search unit 14 searches ingredient data D in descending order of correspondence degree. Ｒ are sorted to extract the first k ingredients, and one or more ingredients that do not belong to the first ingredient group are identified as a second ingredient group.

[0039] In step S137, the second search unit 14 presents the second search result indicating the second group of ingredients to the user. The second search unit 14 transmits the second search result to the user terminal 30. The user terminal 30 receives and displays the second search result. As a result, the user can further confirm the ingredients searched for using the trained model 19.

[0040] A screen for using the search system 10 will be described with reference to FIG. 7 . FIG. 7 shows an example of the screen. In this example, the screen 200 includes an input area 210 for receiving input values ​​for target values, a first display area 220 for displaying first search results, and a second display area 230 for displaying second search results. The input area 210 can receive target values ​​for one or more objective variables, one or more raw material categories, and one or more filters. The first display area 220 includes a table 221 for displaying the first group of raw materials and a table 222 for displaying common characteristics. The second display area 230 is configured with a table for displaying the second group of raw materials. A user viewing the screen 200 can distinguish between raw materials searched using association analysis and raw materials searched using the trained model 19. As in this example, the search system 10 may present the first group of raw materials and the second group of raw materials to the user in a format that visually distinguishes the first group of raw materials from the second group of raw materials.

[0041] As an alternative to screen 200, the search system 10 may present the first and second groups of ingredients to the user in a format that does not visually distinguish between them. For example, the search system 10 may present search results from the two methods in a unified format that mixes the first and second groups of ingredients.

[0042] [Modifications] The technology according to the present disclosure has been described in detail above based on various examples. However, the present disclosure is not limited to the above examples. The technology according to the present disclosure can be modified in various ways without departing from the spirit of the present disclosure.

[0043] The search system according to the present disclosure may not include a model generation unit. That is, a computer system different from the search system may generate the trained model. The search system may identify the second group of ingredients using a trained model in another computer system. Alternatively, a trained model generated by another computer system may be imported into the search system, and the search system may use the trained model.

[0044] In the above example, the search system is constructed as a server in a client-server system. As another example, the search system may be implemented in a stand-alone computer. Alternatively, the search system may be implemented in a user terminal that can access predetermined databases, such as the raw material database and the composition database, via a communication network.

[0045] The processing steps of the method executed by at least one processor are not limited to the above examples. For example, some of the above steps may be omitted, or the steps may be executed in a different order. Furthermore, any two or more of the above steps may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be executed in addition to the above steps.

[0046] In the present disclosure, when comparing the magnitude of two numerical values, either of the two criteria "greater than or equal to" and "greater than" may be used, or either of the two criteria "less than or equal to" and "less than" may be used.

[0047] In the present disclosure, the expression "at least one processor executes a first process, executes a second process, ... executes an nth process" or an expression corresponding thereto indicates a concept including a case where the entity executing the n processes from the first process to the nth process, i.e., the processor, changes midway through. In other words, this expression indicates a concept including both a case where all n processes are executed by the same processor and a case where the processor changes among the n processes according to an arbitrary policy.

[0048] [Supplementary Notes] As can be seen from the various examples above, the present disclosure includes the following aspects: (Supplementary Note 1) A search system comprising at least one processor, wherein the at least one processor: acquires a target value for a composition specified by a user, acquires, for each of a plurality of compositions, composition data indicating one or more ingredients and one or more characteristic values ​​of the composition, acquires, for each of a plurality of ingredients, ingredient data indicating one or more attribute values ​​of the ingredient, performs an association analysis based on the target value, the composition data, and the ingredient data to identify the one or more ingredients corresponding to the target value as a first ingredient group, inputs each of one or more data records of the ingredient data into a trained model that is trained to accept input of data records of the ingredient data and output degrees of correspondence between the ingredients indicated in the data records and the target value, and calculates one or more degrees of correspondence corresponding to the one or more data records, identifies one or more ingredients that correspond to the target value and do not belong to the first ingredient group as a second ingredient group based on the one or more degrees of correspondence, and presents the first ingredient group and the second ingredient group to the user. (Supplementary Note 2) The search system described in Supplementary Note 1, wherein the at least one processor identifies a set of one or more ingredients from the plurality of ingredients that are used in at least one of the plurality of compositions as a used ingredient group based on the composition data and the ingredient data, and performs machine learning using limited ingredient data, which is the ingredient data limited to the used ingredient group, as an explanatory variable and a limited ingredient vector, which is a vector indicating whether each ingredient constituting the used ingredient group is included in the first ingredient group, as a target variable, to generate the trained model.(Supplementary Note 3) The search system according to Supplementary Note 1 or 2, wherein the at least one processor: generates a composition vector based on the composition data, the composition vector being a vector indicating whether each of the plurality of compositions satisfies the target value; generates a composition matrix based on the composition data, for each of the plurality of compositions, the composition matrix being a matrix indicating the raw materials among the plurality of raw materials used for the composition and the raw materials among the plurality of raw materials not used for the composition; and performs the association analysis based on the composition vector and the composition matrix. (Supplementary Note 4) The search system according to any one of Supplements 1 to 3, wherein the at least one processor presents the first group of raw materials and the second group of raw materials to the user in a format in which the first group of raw materials and the second group of raw materials are visually distinguished. (Supplementary Note 5) The search system according to any one of Supplements 1 to 4, wherein the at least one processor: identifies the attribute values ​​common to the first group of raw materials as common features based on the raw material data, and further presents the common features to the user.(Supplementary Note 6) A search method executed by a search system having at least one processor, comprising: a step of acquiring a target value of a composition specified by a user; a step of acquiring, for each of a plurality of compositions, composition data indicating one or more ingredients and one or more characteristic values ​​of the composition; a step of acquiring, for each of a plurality of ingredients, ingredient data indicating one or more attribute values ​​of the ingredient; a step of performing association analysis based on the target value, the composition data, and the ingredient data, to identify one or more ingredients corresponding to the target value as a first ingredient group; a step of inputting each of one or more data records of the ingredient data into a trained model that is trained to accept input of a data record of the ingredient data and output a degree of correspondence between the ingredient indicated in the data record and the target value, and calculating one or more degrees of correspondence corresponding to the one or more data records; a step of identifying one or more ingredients that correspond to the target value and do not belong to the first ingredient group as a second ingredient group based on the one or more degrees of correspondence; and a step of presenting the first ingredient group and the second ingredient group to the user.(Supplementary Note 7) A search program that causes a computer to execute the following steps: acquiring a target value for a composition specified by a user; acquiring, for each of a plurality of compositions, composition data indicating one or more ingredients and one or more characteristic values ​​of the composition; acquiring, for each of a plurality of ingredients, ingredient data indicating one or more attribute values ​​of the ingredient; performing association analysis based on the target value, the composition data, and the ingredient data to identify one or more ingredients that correspond to the target value as a first ingredient group; inputting each of one or more data records of the ingredient data into a trained model that is trained to accept input of a data record of the ingredient data and output the degree of correspondence between the ingredient indicated in the data record and the target value, and calculating one or more degrees of correspondence corresponding to the one or more data records; identifying, based on the one or more degrees of correspondence, one or more ingredients that correspond to the target value and do not belong to the first ingredient group as a second ingredient group; and presenting the first ingredient group and the second ingredient group to the user.

[0049] According to Supplements 1, 6, and 7, ingredients that are expected to achieve the target value of the composition are identified using two techniques, association analysis and a trained model, so that such ingredients can be comprehensively and efficiently presented to the user. In one example, by using these techniques, ingredients that are expected to achieve the target value of the composition can be identified and presented even when some of the characteristic values ​​of the composition data or some of the ingredient data are missing values. As a result, the efficiency of research and development of compositions can be expected to improve.

[0050] According to Supplementary Note 2, a trained model reflecting information about the first raw material group obtained by association analysis is generated by performing machine learning using a target variable that reflects that information. By using a trained model that is dynamically generated in this way, the second raw material group can be identified with higher accuracy.

[0051] According to Appendix 3, the relationship between a composition vector, which indicates whether each of a plurality of compositions satisfies a target value, and a composition matrix, which indicates which raw material is used in which composition, can be obtained by association analysis. By using these vectors and matrices, the accuracy of the association analysis can be improved.

[0052] According to Appendix 4, by presenting groups of ingredients corresponding to the target values ​​separately for each of the two identification methods, the basis for the search (recommendation) of each ingredient presented can be directly or indirectly communicated to the user. Such information can also lead to more efficient research and development of compositions.

[0053] According to Supplementary Note 5, by also displaying the common characteristics of the first candidate group obtained by the association analysis, the basis for the search (recommendation) based on the association analysis can be communicated to the user. Such information can also lead to more efficient research and development of compositions.

[0054] 10...Search system, 11...Acquisition unit, 12...First search unit, 13...Model generation unit, 14...Second search unit, 19...Trained model, 21...Raw material database, 22...Composition database, 30...User terminal, 200...Screen, 210...Input area, 220...First display area, 230...Second display area.

Claims

1. A search system comprising at least one processor, wherein the at least one processor: obtains a target value of a composition specified by a user; obtains composition data indicating one or more raw materials and one or more characteristic values of each of the plurality of compositions; obtains raw material data indicating one or more attribute values of each of the plurality of raw materials; performs an association analysis based on the target value, the composition data, and the raw material data to identify one or more of the raw materials corresponding to the target value as a first raw material group; inputs one or more data records of the raw material data into a learned model trained to receive an input of a data record of the raw material data and output a degree of correspondence between the raw material indicated by the data record and the target value, and calculates one or more of the degrees of correspondence corresponding to the one or more data records; identifies one or more of the raw materials corresponding to the target value and not belonging to the first raw material group as a second raw material group based on the one or more degrees of correspondence; and presents the first raw material group and the second raw material group to the user.

2. The search system according to claim 1, wherein the at least one processor: identifies a set of one or more raw materials used in at least one of the plurality of compositions among the plurality of raw materials as a used raw material group based on the composition data and the raw material data; performs machine learning using the limited raw material data, which is the raw material data limited to the used raw material group, as an explanatory variable and a limited raw material vector, which is a vector indicating whether each raw material constituting the used raw material group is included in the first raw material group, as an objective variable, to generate the learned model.

3. The search system according to claim 1 or 2, wherein the at least one processor: generates a composition vector, which is a vector indicating whether each of the plurality of compositions satisfies the target value, based on the composition data; generates a composition matrix, which is a matrix indicating, for each of the plurality of compositions, the raw materials used for the composition among the plurality of raw materials and the raw materials not used for the composition among the plurality of raw materials, based on the composition data; and performs the association analysis based on the composition vector and the composition matrix.

4. The search system according to claim 1 or 2, wherein the at least one processor presents the first raw material group and the second raw material group to the user in a form that visually distinguishes the first raw material group and the second raw material group.

5. The search system according to claim 1 or 2, wherein the at least one processor: identifies, as common features, the attribute values common in the first raw material group based on the raw material data; and further presents the common features to the user.

6. A search method executed by a search system including at least one processor, the method comprising: obtaining a target value of a composition specified by a user; obtaining, for each of a plurality of the compositions, composition data indicating one or more raw materials and one or more characteristic values of the composition; obtaining, for each of the plurality of the raw materials, raw material data indicating one or more attribute values of the raw material; performing an association analysis based on the target value, the composition data, and the raw material data to identify, as a first raw material group, one or more of the raw materials corresponding to the target value; inputting each of one or more data records of the raw material data into a learned model learned to receive an input of the data record of the raw material data and output a degree of correspondence between the raw material indicated by the data record and the target value, and calculating one or more of the degrees of correspondence corresponding to the one or more data records; identifying, as a second raw material group, one or more of the raw materials corresponding to the target value and not belonging to the first raw material group based on the one or more degrees of correspondence; and presenting the first raw material group and the second raw material group to the user.

7. A search program that causes a computer to execute the steps of: obtaining a target value of a composition specified by a user; obtaining composition data indicating one or more raw materials and one or more characteristic values of each of the plurality of compositions; obtaining raw material data indicating one or more attribute values of each of the plurality of raw materials; performing an association analysis based on the target value, the composition data, and the raw material data to identify one or more of the raw materials corresponding to the target value as a first raw material group; inputting one or more data records of the raw material data into a learned model learned to receive an input of the data records of the raw material data and output a degree of correspondence between the raw materials indicated by the data records and the target value, and calculating one or more of the degrees of correspondence corresponding to the one or more data records; identifying one or more of the raw materials corresponding to the target value and not belonging to the first raw material group as a second raw material group based on the one or more degrees of correspondence; and presenting the first raw material group and the second raw material group to the user.

Images