Machine Learning Methods

Abstract

To provide a machine learning method that considers disclosure or concealment of a test result of a food sample and disclosure or concealment of an analysis result corresponding to the test result.SOLUTION: A second apparatus performs a step of receiving a test result of a food sample by a test device, a step of acquiring an analysis result by analyzing the test result, a step of acquiring first information on disclosure or concealment of the test result and second information on disclosure or concealment of the analysis result, a step of storing the first information and the test result in association with each other and storing the second information and the analysis result in association with each other, and a step of performing machine learning that predicts the analysis result using at least one of the test result and the analysis result, which is set as the disclosure, as teacher data.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present disclosure relates to machine learning methods. [Background technology]

[0002] As disclosed in Patent Document 1, a system for narrowing down search results by inputting multiple search conditions is known as a sake brewing analysis system. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2020-184167 Summary of the Invention [Problem to be solved by the invention]

[0004] Sensory measurement results (hereafter referred to as sensory results) are often used only within each organization, but the more sensory results there are, the more useful they are for research and development. accuracy There is a desire to use sensory results from outside the organization to improve efficiency and efficiency. However, attitudes toward disclosing sensory results vary greatly among private companies and research institutions. The system disclosed in Patent Document 1 did not take into consideration the disclosure or confidentiality of information, including sensory results of sake as a food sample, within a single company or organization.

[0005] Furthermore, in research and development in the food industry, each company or research institute managed the sensory results of food samples within the scope of their own organization, and there was no system for managing sensory results as public information.

[0006] The present disclosure has been made to solve such problems, and its purpose is to provide a machine learning method that takes into account whether sensory results are made public or private. [Means for solving the problem]

[0007] The present disclosure relates to a machine learning method executed by a computing device provided by a system management company, the computing device comprising the steps of: acquiring analysis results and first sensory results of a sample by an analytical instrument; acquiring disclosure setting information regarding whether the analysis results and the first sensory results are to be made public or confidential; storing the disclosure setting information in association with the analysis results and the first sensory results; performing machine learning to predict the first sensory results using at least one of the analysis results set to be public or the first sensory results as training data to generate a trained model; Learn and inputting the second sensory result into the trained model.

[0008] According to the present disclosure, it is possible to provide a machine learning method that takes into consideration whether sensory results are made public or confidential. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram showing the overall configuration of an analysis system according to a first embodiment. [Figure 2] FIG. 2 is a diagram for explaining the flow of the analysis system according to the first embodiment. [Figure 3] FIG. 2 is a diagram showing the configuration of each device in the analysis system according to the first embodiment. [Figure 4] FIG. 2 is a sequence diagram of the analysis system according to the first embodiment. [Figure 5] FIG. 10 is a diagram illustrating learning of an estimation model in a learning phase. [Figure 6] FIG. 10 is a diagram illustrating a configuration of an analysis system in an operation phase. [Figure 7] FIG. 1 shows sensory and analytical results for food samples. [Figure 8] 10 is a diagram showing the number of published analysis results based on customer information, the number of published analysis results, and amount information. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0010] The present embodiment will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference numerals, and their description will not be repeated in principle. <First Embodiment> A machine learning method in an analysis system 1000 according to the first embodiment will be described with reference to Figs. 1 to 4. Fig. 1 is a diagram showing the overall configuration of the analysis system 1000 according to the first embodiment. Fig. 2 is a diagram for explaining the flow of the analysis system 1000 according to the first embodiment. Fig. 3 is a diagram showing the configuration of each device in the analysis system 1000 according to the first embodiment. Fig. 4 is a sequence diagram of the analysis system 1000 according to the first embodiment.

[0011] As shown in FIG. 1, in the analysis system 1000, devices included in a customer 100, a system management company 200, a contract analysis company 300, and a server 400 are connected via a network 5. The customer 100 requests the analysis of food samples, and may be, for example, 1 to N. The system management company 200 manages the entire analysis system 1000, and may be, for example, one. The contract analysis company 300 analyzes food samples, and may be, for example, 1 to M. There may be only one contract analysis company 300. The server 400 stores and publishes various data, and may be, for example, one. There may be multiple servers 400.

[0012] Referring to Figure 2, the flow of the analysis system 1000 between one set of a customer 100, a system management company 200, a contract analysis company 300, and a server 400 will be described. The customer 100 sends a food sample to the contract analysis company 300. The contract analysis company 300 analyzes the food sample using analytical equipment. The contract analysis company 300 notifies the system management company 200 of the analysis results.

[0013] The system management company 200 pays the fee for the analysis request to the contract analysis company 300. The system management company 200 has its sensory measurement evaluators conduct sensory tests on food samples. Sensory tests are tests that use human senses (sight, hearing, taste, smell, touch, etc.) to determine the quality of a product, and are used for foods, fragrances, industrial products, etc. Sensory results are obtained from sensory tests.

[0014] The system management company 200 acquires the analysis results. The system management company 200 analyzes the acquired analysis results and acquires the analysis results, which are the first sensory results or the second sensory results. The analysis results are, for example, review results created based on the analysis results and the sensory results. The system management company 200 notifies the customer 100 of the analysis results.

[0015] The customer 100 sets whether the analysis results should be made public or kept secret, sets whether the analysis results should be made public or kept secret, and notifies the system management company 200 of the selection result. The system management company 200 transmits amount information to the customer 100. The customer 100 pays the amount to the system management company 200 based on the amount information. The system management company 200 transmits the public data to be made public based on the selection result to the server 400.

[0016] 3, the configuration of various devices used in the analysis system 1000 will be described. The analysis system 1000 includes a first device 1 provided by the customer 100, a second device 2 provided by the system management company 200, a third device 3 provided by the contract analysis company 300, and a server device 4 provided by the server 400, all connected via a network 5.

[0017] The first device 1, the second device 2, and the third device 3 are configured according to a general-purpose computer architecture. In this embodiment, the first device 1, the second device 2, and the third device 3 are configured as desktop computers. The first device 1, the second device 2, and the third device 3 may be devices other than desktop computers, such as laptop computers, tablet computers, and mobile terminals such as smartphones.

[0018] The server device 4 is configured in accordance with a general-purpose computer architecture. In this embodiment, the server device 4 is owned by a system management company 200.

[0019] The first device 1 includes a processor 11, a main memory 12, an input / output interface 13, a communication interface 14, and a storage 15. These components are connected via a bus.

[0020] The processor 11 is a computing entity (computer) that executes various processes according to various programs. The processor 11 is configured with at least one of a CPU (Central Processing Unit), an FPGA (Field Programmable Gate Array), a GPU (Graphics Processing Unit), and an MPU (Multi-Processing Unit), for example. The processor 11 may also be configured with a processing circuitry. The processor 11 reads out programs stored in the storage 15, expands them into the main memory 12, and executes them.

[0021] The main memory 12 may be a volatile storage device such as a random access memory (RAM), a dynamic RAM (DRAM), or a static random access memory (SRAM), or a non-volatile storage device such as a read only memory (ROM). memory It consists of a device.

[0022] The input / output interface 13 transmits input signals from the user via buttons, touch panels, etc., and output signals to display devices such as liquid crystal displays, to each device.

[0023] The communication interface 14 transmits and receives data (information) to and from other devices via wired or wireless connection. In this embodiment, the communication interface 14 transmits and receives data (information) to and from other communication interfaces via wireless communication via the network 5.

[0024] The storage 15 is configured with, for example, a nonvolatile storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The storage 15 stores analytical information 151 including sensory evaluation items of the sensory test of the food sample and analytical evaluation items corresponding to the sensory evaluation items, selection information 152 which is information selecting data to be made public and data to be kept confidential, and the like.

[0025] The second device 2 includes a processor 21, a main memory 22, an input / output interface 23, a communication interface 24, and a storage 25. These components are connected via a bus.

[0026] The configuration of the processor 21, main memory 22, input / output interface 23, communication interface 24, and storage 25 in the second device 2 is similar to the configuration of the processor 11, main memory 12, input / output interface 13, communication interface 14, and storage 15 in the first device 1.

[0027] The storage 25 stores customer information 251, ID information 252, an estimation model 253, analysis result information 254, analysis result information 255, public / confidential information 256, and the like.

[0028] Customer information 251 is information about customer 100. ID information 252 is information for identifying the customer that is associated with customer information 251. Estimation model 253 is a model used for machine learning, and will be described in detail below. Analysis result information 254 is information about the analysis results of the food sample. Analysis result information 255 is information about the first sensory result or the second sensory result created based on analysis result information 254 and the sensory result. Public / confidential information 256 is information indicating which items of analysis result information 254 and analysis result information 255 are set to public or confidential.

[0029] The third device 3 includes a processor 31, a main memory 32, an input / output interface 33, a communication interface 34, a storage 35, and an analytical instrument interface 36. These components are connected via a bus.

[0030] The configuration of the processor 31, main memory 32, input / output interface 33, communication interface 34, and storage 35 in the third device 3 is similar to the configuration of the processor 11, main memory 12, input / output interface 13, communication interface 14, and storage 15 in the first device 1.

[0031] The analytical instrument interface 36 inputs and outputs information to and from analytical instruments for analyzing food samples that are provided by the contract analysis company 300. The storage 35 stores analysis result information 351 relating to the analysis results of food samples, etc.

[0032] The server device 4 includes a processor 41, a main memory 42, a communication interface 44, and a storage 45. These components are connected via a bus.

[0033] The configurations of the processor 41, main memory 42, communication interface 44, and storage 45 in the server device 4 are similar to the configurations of the processor 11, main memory 12, communication interface 14, and storage 15 in the first device 1.

[0034] The storage 45 stores analysis result information 454, analysis result information 455, and public / confidential information 456 corresponding to the analysis result information 254, analysis result information 255, and public / confidential information 256 transmitted from the second device 2.

[0035] 4, the flow of processing between the first device 1 of the customer 100, the second device 2 of the system management company 200, the third device 3 of the contracted analysis company 300, and the server device 4 of the server 400 will be described. "SQ" in FIG. 4 stands for sequence.

[0036] In SQ11, the first device 1 transmits the analysis information to the third device 3. In SQ12, the third device 3 analyzes the food sample using an analytical instrument based on the analysis item information contained in the received analysis information. Thereafter, in SQ12, the third device 3 creates analysis result information. In SQ13, the third device 3 transmits the analysis result information to the second device 2.

[0037] In SQ14, the second device 2 registers the received analysis result information, analyzes the analysis result information, and acquires analysis result information as the first sensory result. The analysis result information includes any of sensory result information created based on the results of the sensory test obtained from the sensory measurement evaluator, sensory result prediction information obtained by machine learning using the estimation model 253, and consideration result information obtained from the analysis result information and the sensory result information (or sensory prediction result information).

[0038] In SQ15, the second device 2 transmits the analysis result information and the analysis result information to the first device 1. The analysis result information and the analysis result information are associated with an ID for identifying the customer. In SQ16, the first device 1 sets whether the received analysis result information and the analysis result information should be made public or kept confidential.

[0039] In SQ17, the first device 1 transmits selection information indicating whether to make the analysis result information and analytical result information public or confidential to the second device 2. In SQ18, the second device 2 tags each item of the analytical result information and each item of the analytical result information as public or confidential. In SQ19, the second device 2 is configured to perform machine learning to predict analytical result information using the analytical result information and analytical result information tagged as public as training data. Specifically, a trained model is generated by performing machine learning. Then, by inputting the analytical results set as public or confidential into the trained model, analytical result information can be obtained as a second sensory result.

[0040] In SQ20, the second device 2 is a public Ta In SQ21, the second device 2 transmits the tagged analysis result information and analysis result information to the server device 4. In SQ21, the second device 2 transmits the public or private information linked to the customer information. Ta In SQ21, the second device 2 transmits billing amount information based on the value calculated in SQ21 to the first device 1. The billing amount may include, for example, a usage fee for the machine learning method, a flat rate charged as a subscription, an analysis fee for obtaining analysis result information, etc.

[0041] The learning phase and operation phase of estimation model 253 of this embodiment will be described with reference to Fig. 5 and Fig. 6. Fig. 5 is a diagram for explaining learning of the estimation model in the learning phase. Fig. 6 is a diagram showing the configuration of the analysis system in the operation phase.

[0042] 5, the learning phase is a pre-learning phase in which the estimation model 253 is trained before the analysis result or the analysis result and the sensory result are provided to the second device 2. The estimation model 253 is trained by the learning device 51 to estimate the sensory result as the second sensory result from the analysis result, or to predict the examination result as the second sensory result from the analysis result and the sensory result.

[0043] Known algorithms such as supervised learning, unsupervised learning, and reinforcement learning can be used as a learning algorithm for training the estimation model 253. In this embodiment, the learning device 51 trains the estimation model 253 by supervised learning using the training data 40.

[0044] The training data 40 is prepared in advance for training the estimation model 253, and includes analysis results and sensory results corresponding to the analysis results. For example, the program designer associates multiple past analysis results of food samples with the sensory results (correct answer data) corresponding to the analysis results to create the training data (teacher data) 40. The designer prepares multiple pieces of such training data 40 in advance.

[0045] The estimation model 253 includes a neural network 2531 and parameters 2532 used by the neural network 2531. The neural network 2531 may be a known neural network used in deep learning processing, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a long short-term memory network (LSTM).

[0046] The estimation model 253 performs deep learning by using the neural network 2531 described above. The parameters 2532 include weighting coefficients and the like used in calculations by the neural network 2531. Note that the estimation model 253 is not limited to one that learns by deep learning using a neural network, and may also learn by other machine learning methods. Here, the pre-learning estimation model 253 and the trained estimation model 253 are collectively referred to as "estimation models," while the trained estimation model 253 is also particularly referred to as "trained model."

[0047] The learning device 51 receives input of analysis results or input of analysis results and sensory results from the learning data 40. The learning device 51 executes processing to estimate the sensory results or the examination results based on the input analysis results or the analysis results and sensory results and the estimation model 253 including the neural network 2531.

[0048] The learning device 51 trains the estimation model 253 based on the estimation result of the sensory result and the correct answer data (sensory result corresponding to the analysis result) included in the learning data 40. Specifically, the learning device 51 trains the estimation model 253 by adjusting the parameters 2532 (for example, weighting coefficients) so that the estimation result of the sensory result obtained by the estimation model 253 approaches the correct answer data, thereby generating a trained model.

[0049] Alternatively, the learning device 51 trains the estimation model 253 based on the estimation result of the examination result and the correct answer data (examination result corresponding to the analysis result and the sensory result) included in the learning data 40. Specifically, the learning device 51 trains the estimation model 253 by adjusting the parameters 2532 (for example, weighting coefficients) so that the estimation result of the examination result obtained by the estimation model 253 approaches the correct answer data, thereby generating a trained model.

[0050] The operation phase shown in FIG. 6 is a phase in which, after providing an analysis result or an analysis result and an analysis result, a sensory prediction result or a study result is estimated as a second sensory result using an estimation model 253, which is a trained model. As shown in FIG. 6, the second device 2 stores the estimation model 253 trained by the learning device 51 shown in FIG. 5 in the storage 25. For example, the second device 2 acquires the estimation model 253 from the learning device 51 and stores the acquired estimation model 253 in the storage 25. Note that the learning device 51 may be the second device 2, and the functions of the learning device 51 described above may be functions possessed by the processor 21 of the second device 2.

[0051] The processor 21 of the second device 2 includes an input unit 211, a processing unit 212, and an output unit 213. The input unit 211 accepts input of an analysis result or an analysis result and an analysis result. The processing unit 212 executes processing to estimate a sensory prediction result or an examination result based on the analysis result or the analysis result and an analysis result input from the input unit 211 and an estimation model 253 including a neural network 2531. As described above, the estimation model 253 is not limited to one that learns by deep learning using a neural network, but may also learn by other machine learning methods. The output unit 213 outputs the result obtained by the processing unit 212 as a sensory prediction result or an examination result.

[0052] FIG. 7 shows the sensory and analytical results of food samples. trial Specific examples of sensory and analytical results are shown for a case where multiple sakes were used as ingredients. In the case of sake, for example, a sensory evaluator conducts a sensory test to quantify the sensory evaluation items of "intensity of aroma," "strongness of flavor," "sweetness / dryness," and "ripeness." Analytical items corresponding to the sensory evaluation items, such as "alcohol content," "sake meter value," "acidity," "amino acid content," "isoamyl acetate," "isoamyl alcohol," "ethyl caproate," "citric acid," "pyruvic acid," "malic acid," "succinic acid," "lactic acid," "acetic acid," and "glucose," are analyzed and quantified by analytical equipment.

[0053] More specifically, for example, the amounts of the analytical items "ethyl caproate" and "glucose" affect the sensory evaluation item "fragrance strength." If the amounts of these are high, it can be determined that the "fragrance strength" is strong.

[0054] The second device 2 creates an examination result showing the relationship between the analytical result and the sensory result based on the analytical result and the sensory result shown in Fig. 7. The examination result is, for example, data showing that a certain peak value in the chromatogram of the analytical result corresponds to a certain taste item in the sensory evaluation.

[0055] FIG. 8 is a diagram showing the number of published analysis results, the number of published analysis results, and monetary amount information based on customer information. Customer 100 sets whether each item of the acquired analysis results will be made public or confidential, and also sets whether each item of the acquired analysis results will be made public or confidential. Information about the analysis results and whether the analysis results will be made public or confidential is also referred to as disclosure setting information. The disclosure setting information includes first information about whether the analysis results will be made public or confidential, and second information about whether the analysis results (first sensory results and second sensory results) will be made public or confidential.

[0056] The second device 2 of the system management company 200 calculates the amount to be billed to the customer 100 based on the public or private disclosure setting information transmitted from the first device 1 of the customer 100. As shown in FIG. 8, for example, customer A sets the first information to make 14 of a total of 14 analysis result items public, and sets the second information to make 4 of a total of 4 analysis result items public. In this case, the discount rate on the billed amount for customer A is 15%. Customer B sets the first information to make 10 of a total of 14 analysis result items public, and sets the second information to make 3 of a total of 4 analysis result items public. In this case, the discount rate on the billed amount for customer B is 7%.

[0057] Customer C sets the first information to disclose 8 of the 14 total analysis result items, and sets the second information to disclose 2 of the 4 total analysis result items. In this case, the discount rate on Customer C's bill is 5%. Customer D sets the first information to disclose 5 of the 14 total analysis result items, and sets the second information to disclose 1 of the 4 total analysis result items. In this case, the discount rate on Customer D's bill is 3%. Customer E sets the first information to disclose 0 of the 14 total analysis result items, and sets the second information to disclose 0 of the 4 total analysis result items. In this case, the discount rate on Customer E's bill is 0%.

[0058] As shown in Figure 8, the discount rate for the amount to be charged is set to differ depending on the number of published analysis results and the number of published analysis results. The more published results, the higher the discount rate for the amount to be charged. This has the advantage that the party that publishes the results can receive a larger discount on the amount to be charged by publishing more results.

[0059] In the machine learning method of this embodiment, customer 100 can select whether to make public or confidential the analysis results and each item of the analysis results obtained. Second device 2 of system management company 200 tags the results as public or confidential, and is configured to perform machine learning using the analysis result information and analysis result information tagged as public as training data.

[0060] In this way, accuracy and efficiency can be improved by using analysis result information and analysis result information tagged as public. Since the discount rate on the billing amount varies depending on the number of analysis result information and analysis result information tagged as public, it is possible to encourage customers 100 to actively disclose their information. <Modification> In the above-described embodiment, the discount rate on the billing amount was set to vary depending on the number of items published. However, the discount rate on the billing amount may also be set to vary depending on the content of the items to be published. For example, the discount rate may be set to be higher for customers who have published items with fewer published data than for other customers.

[0061] In the above-described embodiment, the processor 21 of the second device 2 outputs a sensory prediction result from the input analysis result. The processor 21 may output a consideration result from the input analysis result and the sensory evaluation result, or may output a consideration result from the input analysis result and the sensory prediction result.

[0062] In the above embodiment, the sensory test is performed by the system management company 200. However, the sensory test may be performed by the customer 100 or a contract analysis company 300 that has analytical equipment. For example, if the sensory test is performed by the customer 100, the billing amount may be reduced accordingly.

[0063] In the above-described embodiment, the third device 3 of the contract analysis company 300 may transmit the analysis results directly to the first device 1 of the customer 100, rather than to the second device 2 of the system management company 200. In this way, the processes of receiving and transmitting the analysis results by the second device 2 of the system management company 200 can be omitted.

[0064] In the above-described embodiment, the scope of disclosure that can be referenced may be set for each customer. For example, if customer A and customer B are related companies, the scope of disclosure that can be referenced may be wide, and if customer A and customer C are competing companies, the scope of disclosure that can be referenced may be narrower than the relationship with customer B.

[0065] In the above-described embodiment, instead of changing the amount charged depending on the disclosure status of the information, the search range of information may be set to be wide when the number of disclosures is large and narrow when the number of disclosures is small.

[0066] In the above-described embodiment, it may be possible for companies other than the requesting customer to search for information on the server by paying a fee.

[0067] In the above-described embodiment, the analytical information transmitted from first apparatus 1 to second apparatus 2 may include information such as the production date of the food sample and the type of food sample.

[0068] [Aspect] It will be appreciated by those skilled in the art that the exemplary embodiments described above are examples of the following aspects.

[0069] (Item 1) A machine learning method according to one aspect is a machine learning method executed by a computing device provided by a system management company. The computing device includes steps of acquiring analysis results and first sensory results of a sample by an analytical instrument, acquiring disclosure setting information regarding whether the analysis results and the first sensory results are to be made public or confidential, storing the disclosure setting information in association with the analysis results and the first sensory results, performing machine learning to predict the first sensory results using at least one of the analysis results set to be public or the first sensory results as training data to generate a trained model, and generating a trained model based on the analysis results set to be confidential. Learn and inputting the second sensory result into the trained model.

[0070] According to the machine learning method described in paragraph 1, it is possible to provide a machine learning method that takes into consideration whether sensory results are made public or confidential.

[0071] (Section 2) The first sensory result is either the sensory evaluation result of the sample, the sensory prediction result output by the calculation device based on the analysis result, or the examination result that is associated from the analysis result and the sensory evaluation result or the analysis result and the sensory prediction result.

[0072] According to the machine learning method described in paragraph 2, it is possible to provide a machine learning method that takes into consideration whether any of the sensory evaluation results, sensory prediction results, and examination results are made public or confidential.

[0073] (Item 3) The sensory evaluation results are the results of the evaluation of the sample by a sensory evaluator.

[0074] According to the machine learning method described in paragraph 3, the sensory evaluation results evaluated by the sensory measurement evaluator can be set to be public or confidential.

[0075] (4) The calculation device executes a step of calculating the input analysis results and outputting the sensory prediction results.

[0076] According to the machine learning method described in Section 4, a sensory prediction result can be output from the analysis result, thereby improving the accuracy of the sensory prediction result.

[0077] (Item 5) The calculation device executes a step of calculating the input analysis result and the input sensory evaluation result or the input sensory prediction result, and outputting the examination result.

[0078] According to the machine learning method described in paragraph 5, the examination result can be output from the analysis result and the sensory evaluation result or the sensory prediction result, so that the accuracy of the examination result can be improved.

[0079] (Section 6) The disclosure setting information is stored in association with the customer information of the system management company.

[0080] According to the machine learning method described in Section 6, it is possible to appropriately manage information to be made public or kept secret for each piece of customer information.

[0081] (Clause 7) The computing device executes a step of changing billing amount information to be billed to the customer in accordance with the disclosure setting information associated with the customer information.

[0082] According to the machine learning method described in Section 7, it is possible to charge an appropriate amount depending on the information that the customer chooses to disclose or conceal.

[0083] (Clause 8) The computing device executes a step of publishing the analysis results, first sensory results, and second sensory results that are set to be public in the public setting information on the server via the network, and setting the analysis results, first sensory results, and second sensory results that are set to be confidential not to be published on the server.

[0084] According to the machine learning method described in paragraph 8, only the analysis results, first sensory results, and second sensory results that are set to be made public can be appropriately made public on the server.

[0085] (Article 9) The disclosure setting information includes first information regarding whether the analysis results are to be made public or kept secret, and second information regarding whether the first sensory result and the second sensory result are to be made public or kept secret.

[0086] According to the machine learning method described in paragraph 9, the first information and the second information can be set individually.

[0087] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present disclosure is defined by the claims, not by the description of the above embodiments, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0088] 1 first device, 2 second device, 3 third device, 4 server device, 5 network, 11, 21, 31, 41 processor, 12, 22, 32, 42 main memory, 13, 23, 33 input / output interface, 14, 24, 34, 44 communication interface, 15, 25, 35, 45 storage, 36 analytical equipment interface, 40 learning data, 51 learning device, 151 analysis information, 152 selection information, 200 system management company, 211 input unit, 212 processing unit, 213 output unit, 251 customer information, 252 ID information, 253 estimation model, 254, 351, 454 analysis result information, 255, 455 analysis result information, 256, 456 public / confidential information, 300 contract analysis company, 400 server, 1000 analysis system, 2531 Neural Network, 2532 parameters.

Claims

1. A machine learning method executed by a computing device provided by a system management company, The computing device obtaining an analytical result of the sample by the analytical instrument and a first sensory result; acquiring disclosure setting information regarding disclosure or confidentiality of the analysis result and the first sensory result; storing the public setting information in association with the analysis result and the first sensory result; a step of performing machine learning to predict the first sensory result using at least one of the analysis result and the first sensory result set to be public as training data to generate a trained model; A machine learning method that executes a step of inputting the analysis result, which has been set to be confidential, into the trained model and outputting a second sensory result.

2. 2. The machine learning method according to claim 1, wherein the first sensory result is one of a sensory evaluation result of the sample, a sensory prediction result output by the computing device based on the analysis result, and a study result associated with the analysis result and the sensory evaluation result or the analysis result and the sensory prediction result.

3. The machine learning method according to claim 2 , wherein the sensory evaluation result is a result of evaluation of the sample by a sensory measurement evaluator.

4. The machine learning method according to claim 2 or 3, wherein the arithmetic device executes a step of calculating the input analysis result and outputting the sensory prediction result.

5. The machine learning method according to claim 2 , wherein the arithmetic device performs a step of calculating the input analysis result and the input sensory evaluation result or the input sensory prediction result, and outputting the examination result.

6. The machine learning method according to claim 1 , wherein the public setting information is stored in association with customer information of the system management company.

7. The machine learning method according to claim 6 , wherein the computing device executes a step of changing billing amount information to be billed to the customer in accordance with the disclosure setting information associated with the customer information.

8. The machine learning method according to any one of claims 1 to 7, wherein the computing device executes a step of publishing the analysis results, the first sensory results, and the second sensory results that are set to be public in the public setting information on a server via a network, and setting the analysis results, the first sensory results, and the second sensory results that are set to be confidential not to be published on the server.

9. The public setting information is First information regarding whether the analysis results will be made public or confidential; The machine learning method according to claim 1 , further comprising: second information regarding the first sensory result and whether the second sensory result is made public or private.

Images