JP2022127316A - Physical property predictor and program - Google Patents

Abstract

To provide a physical property predictor and a program allowing a user to learn, by each prediction, a degree of probability of a physical property prediction.SOLUTION: Learning-purpose condition data storage means 20 stores multiple pieces of learning-purpose condition data each indicating a generation condition of a substance. A machine learning model 40 outputs predicted physical property data learned through the multiple pieces of learning-purpose condition data and indicating properties of a substance to be generated at a generation condition shown by input condition data, when inputting the prescribed input condition data. A similarity determination part 50 determines a similarity between each of the multiple pieces of learning-purpose condition data and the input condition data. A prediction physical property data output part 60 outputs the predicted physical property data and also outputs information based on one of similarities as a likelihood ratio of the predicted physical property data.SELECTED DRAWING: Figure 2

Description

Patent Law Article 30, Paragraph 2 application filed February 15, 2021, https://www. y-yokohama. com, https://www. y-yokohama. com/release/, https://www. y-yokohama. com/release/? id=3517, https://www. y-yokohama. com/release/pdf/2021021515tr001. pdf

The present invention relates to a physical property prediction device and program.

In Patent Document 1 below, in order to predict the physical properties (safety such as toxicity) of an unknown sample that is a compound, a sample whose degree of similarity with an input unknown sample exceeds a specified value (a plurality of known samples ), create a sub-sample set from the retrieved samples, and perform data analysis on this sub-sample set. In addition, in Patent Document 2 below, a base model and a correction model (a model that predicts the reciprocal of the residual of the base model) are searched for a representative vector close to an unknown input vector that is data of an unknown sample, and the base model and It is described that a physical property predicted value based on the predicted value of the correction model is calculated, and a risk value for the physical property predicted value is calculated based on the predicted value of the correction model.

Japanese Patent No. 5083320 Japanese Patent Application Laid-Open No. 2020-187417

The certainty of physical property prediction differs for each prediction. If it is possible to know the degree of certainty for each prediction, the user can use it to determine which prediction to rely on when making a prototype.

An object of the present disclosure is to provide a physical property prediction device and a program capable of knowing the degree of certainty of physical property prediction for each prediction.

A physical property prediction apparatus according to the present disclosure includes learning condition data storage means for storing a plurality of learning condition data respectively indicating substance production conditions; parameter storage means for storing parameters of a machine learning model for outputting predicted physical property data indicating properties of a substance produced under the production conditions indicated by the input condition data when inputting a and the input condition data, similarity determination means for determining similarity between the input condition data and output means for outputting the predicted physical property data and outputting information based on any of the similarities as the accuracy of the predicted physical property data and including. According to this, the degree of certainty of physical property prediction can be known for each prediction.

A program according to the present disclosure includes learning condition data storage means for storing a plurality of learning condition data respectively indicating substance production conditions, learning by the plurality of learning condition data, and input of given input condition data. parameter storage means for storing parameters of a machine learning model for outputting predicted physical property data indicating properties of a substance produced under the production conditions indicated by the input condition data, each of the plurality of learning condition data, and the input A program for causing a computer to function as similarity determination means for determining similarity between condition data and output means for outputting the predicted physical property data and outputting the similarity as the accuracy of the predicted physical property data is. According to this, a computer can be used to know the degree of certainty of physical property prediction for each prediction.

1 is a diagram showing the configuration of a physical property prediction device that is an example of an embodiment of the present disclosure; FIG. It is a functional block diagram which shows an example of the function implemented in a physical-property prediction apparatus. It is a figure which shows an example of several data for learning memorize|stored in the data storage part for learning. It is a figure which shows an example of several input data. It is a figure which shows an example of the machine learning which a learning part performs. It is a figure which shows an example of the prediction using the machine-learning model which a prediction part performs. It is a figure which shows an example of the prediction result with respect to input data. It is a figure which shows an example of the similarity determination method of the data for learning, and input data. FIG. 10 is a diagram showing another example of prediction results for input data; It is a figure which shows an example of the flow of the prediction process performed with a physical-property prediction apparatus. FIG. 10 is a diagram showing an example of attribute data generated for attributes of condition item data; FIG. 10 is a diagram showing an example of attribute data generated for attributes of condition item data; FIG. 10 is a diagram showing an example of attribute data generated for attributes of condition item data;

Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the present embodiment, by inputting input data indicating conditions for material generation and input (conditions related to raw materials, generation process, physical property measurement, etc.) into a trained machine learning model, the properties of the material to be generated along with outputting predicted physical property data indicating the accuracy of the predicted physical property data. Below, an example in which the substance is a mixture (for example, a tire material) produced by a plurality of raw materials and a production process will be described, but the substance is not limited to a mixture of a plurality of raw materials, and Anything that can be generated is acceptable.

[1. Hardware configuration]
FIG. 1 is a diagram showing the configuration of a physical property prediction device 10 that is an example of an embodiment of the present disclosure. The physical property prediction apparatus 10 according to this embodiment is a computer such as a personal computer, a general-purpose computer, or a personal digital assistant, and as shown in FIG. 15 included. The physical property prediction apparatus 10 may include an optical disk drive for reading optical disks, a USB (Universal Serial Bus) port, and the like.

The processor 11 is a program control device such as a CPU (Central Processing Unit) that operates according to a program installed in the physical property prediction apparatus 10, which is a computer, for example. The storage unit 12 is a storage element such as a ROM (Read Only Memory) or a RAM (Random Access Memory), a hard disk drive, or the like. The storage unit 12 stores data such as programs to be executed by the processor 11 . The communication unit 13 is, for example, a communication interface such as a network board. The display unit 14 is a display device such as a liquid crystal display, and displays various images according to instructions from the processor 11 . The operation unit 15 is a user interface such as a keyboard and a mouse, and receives user operation input and outputs a signal indicating the content of the input to the processor 11 .

[2. Function block]
The physical property prediction device 10 outputs predicted physical property data indicating, for example, the performance predicted for the properties of a substance that is a mixture (for example, the hardness of the substance, the durability and rolling resistance when used as a tire, etc.), Output information about the accuracy of the predicted physical property data. The output of predicted physical property data and accuracy information by the physical property prediction device 10 will be described below.

FIG. 2 is a functional block diagram showing an example of functions implemented in the physical property prediction device 10. As shown in FIG. As shown in FIG. 2, the physical property prediction apparatus 10 functionally includes a learning data storage unit 20 (learning condition data storage unit), an input data acquisition unit 30 (input condition data acquisition unit), and a machine learning It includes a model 40 , a similarity determination section 50 , a predicted physical property data output section 60 and an item name output section 70 . Note that the physical property prediction apparatus 10 does not have to implement all the functions shown in FIG. 2, and may implement functions other than the functions shown in FIG.

[2-1. Learning data storage unit]
The learning data storage unit 20 stores a plurality of learning condition data. “Learning condition data” is data indicating conditions for generating and measuring substances, and is teacher data used for learning of the machine learning model 40 described later. Hereinafter, the learning condition data is also simply referred to as “learning data”. The learning data storage unit 20 may be realized mainly by the storage unit 12, but is not limited to this, and may be an external storage device or a NAS (Network Attached Storage) connected to the physical property prediction device 10 by wire or wirelessly. It may be implemented by other storage devices.

FIG. 3 is a diagram showing an example of a plurality of learning data stored in the learning data storage unit 20. As shown in FIG. In the example shown in FIG. 3, data (data for one row in the table) specified by individual IDs (1, 2, N, etc.) corresponds to one piece of learning data.

The learning data includes a plurality of condition item data as data for each of a plurality of items indicating conditions for generating and measuring substances. Each condition item data is data of an item relating to the conditions for generating a substance or the conditions for measurement. In the example shown in FIG. 3, the learning data includes, as multiple condition item data, values of multiple raw materials (raw material 1 to raw material R), values of multiple process conditions (process 1 to process condition S), and multiple and the values of the measurement conditions (measurement condition 1 to measurement condition T).

"Raw material" is a raw material contained in a substance that is a mixture, and "raw material data" is data such as a value indicating the degree of raw material contained in a substance (mass, weight, blending ratio, etc.). For example, in the learning data identified by ID "1", the value of raw material 1 (first raw material) is "44", the value of raw material 2 (second raw material) is "10", The value of raw material P (P-th raw material) is "7", the value of raw material P+1 (P+1-th raw material) is "3", and the value of raw material Q (Q-th raw material) is "1". , raw material Q+1 (Q+1st raw material) has a value of "2", and raw material R (Rth raw material) has a value of "3". Further, in the learning data identified by the ID of "2", the value of raw material 1 is "43", the value of raw material 2 is "10", the value of raw material P is "6", and the value of raw material P is "6". It shows that the value of P+1 is "3", the value of material Q is "1", the value of material Q+1 is "2", and the value of material R is "0".

"Process conditions" are the conditions for processing (processing, etc.) when producing substances from raw materials, and "data of process conditions" are the conditions for processing substances (for example, when heating or cooling). (temperature and time). For example, in the learning data identified by ID "1", the value of process condition 1 (first process condition) is "148" and the value of process condition S (Sth process condition) is "189". ” is shown. Further, in the learning data identified by ID "2", the value of process condition 1 is "147" and the value of process condition S is "178".

"Measurement conditions" are the conditions for measuring the physical properties of the generated substance, and "measurement condition data" are the data indicating the conditions for measuring the substance (e.g. measurement method and measurement time). is. For example, in learning data identified by an ID of "1", the value of measurement condition 1 (first measurement condition) is "2", and the value of measurement condition 2 (second measurement condition) is "3". ” is shown. Also, in the learning data identified by ID "2", the value of measurement condition 1 is "2" and the value of measurement condition 2 is "3".

The reason why the measurement condition data is included in the condition item data of the learning data is that the data measured as physical properties (physical property values in FIG. 3) are different as the measurement conditions of the generated substances are different. is.

In this embodiment, one of a plurality of attributes is assigned to each condition item data of the learning data. In the example shown in FIG. 3, five attributes are assigned to the learning data. More specifically, among the plurality of condition item data included in the learning data, "attribute 1" is assigned to raw materials 1 to P, and "attribute 2" is assigned to raw materials 1 and raw materials P+1 to raw materials Q. "Attribute 3" is assigned to raw materials Q+1 to R, "attribute 4" is assigned to process conditions 1 to S, and "attribute 5" is assigned to measurement conditions 1 to T. It is In the example shown in FIG. 3, according to the type of raw material condition item data (raw material 1 to raw material R) (for example, whether it is a polymer material, filler, or other) , attribute 1 to attribute 3 are assigned. In the example shown in FIG. 3, one of attributes 1 to 3 is assigned to a plurality of raw material data. Other attributes than 3 may be used.

The learning data also includes physical property data indicating the physical properties (hardness, etc.) of substances generated under a plurality of conditions (conditions indicated by a plurality of condition item data) included in the learning data. In the example shown in FIG. 3, the physical property data are physical property values and indicated by numerical values. For example, learning data identified by an ID of "1" has a physical property value of "1.2", and learning data identified by an ID of "2" has a physical property value of "1.3". It is shown. It should be noted that the physical property data is not limited to numerical values such as physical property values, and may be anything as long as the physical property can be identified. The physical property data may be represented by a sequence of numbers, or may be represented by symbols or characters (hard, soft, etc.).

As shown in FIG. 3, the plurality of condition item data included in each of the plurality of learning data are different among the learning data. For this reason, the physical property values of the learning data are basically different from each other, but there are cases where the physical property values of a plurality of learning data match due to rounding processing such as rounding off the physical property values.

[2-2. Input data acquisition unit]
The input data acquisition unit 30 acquires input condition data to be input to the machine learning model 40, which will be described later. "Input condition data" means data indicating the conditions for the production and measurement of substances, and is data created by the user in order to predict the physical properties (hardness, etc.) of substances produced and measured under the relevant conditions. be. Below, the input condition data is also simply referred to as "input data". The input data acquisition unit 30 may be implemented mainly by the processor 11 of the physical property prediction device 10, or may be implemented by a processor of another information processing device connected to the physical property prediction device 10 via a network. .

FIG. 4 is a diagram showing an example of a plurality of input data created by a user. In the example shown in FIG. 4, as in FIG. 3, data specified by individual IDs (1, 2, M, etc.) (one row of data in the table) corresponds to one piece of input data.

As with the learning data, the input data also includes a plurality of condition item data as data for each of a plurality of items indicating conditions for generating and measuring substances. In the example shown in FIG. 4, as in FIG. 3, the input data includes data of a plurality of raw materials (raw material 1 to raw material R) and a plurality of process conditions (process 1 to process condition S) as a plurality of condition item data. and data of a plurality of measurement conditions (measurement condition 1 to measurement condition T). Since the content of each condition item data is the same as that of the learning data, the explanation is omitted. A user can create input data by, for example, inputting values of condition item data using the operation unit 15 of the physical property prediction apparatus 10 .

Also, like the condition item data included in the learning data, each condition item data of the input data is assigned one of a plurality of attributes. Since the content of the attribute assigned to each condition item data is also the same as that of the learning data, the description is omitted.

As described above, the input data is for predicting the physical properties (hardness, etc.) of the substance generated and measured by the condition item data included therein. Therefore, unlike the learning data, the input data does not include physical property data such as physical property values. By creating input data including a plurality of generations and/or conditions and inputting them to the machine learning model 40, which will be described later, the user can predict physical properties predicted by the machine learning model 40 (more specifically, the prediction unit 43). information can be obtained.

[2-3. machine learning model]
The machine learning model 40 is a model learned by a plurality of learning data stored in the learning data storage unit 20, and is a model that outputs predicted physical property data described later when given input data is input. is. As shown in FIG. 3, the machine learning model 40 includes a parameter storage unit 41, a learning unit 42, and a prediction unit 43. The machine learning model 40 is a model machine-learned by a given neural network such as DNN (Deep Neural Network) or CNN (Convolutional Neural Network). Alternatively, for example, the machine learning model 40 may be a model machine-learned by a given statistical method, linear regression, or the like.

[2-3-1. Parameter storage section]
The parameter storage unit 41 stores parameters of the machine learning model 40 . More specifically, the parameter storage unit 41 stores parameters of the machine learning model 40 that outputs predicted physical property data described later when given input data is input to the machine learning model 40 . Note that the parameter storage unit 41 may be implemented mainly by the storage unit 12 of the physical property prediction device 10, or may be implemented by another storage device or the like.

When the machine learning model 40 is realized by a neural network, the parameter storage unit 41 stores a plurality of nodes constituting the neural network, the weighting of each node, the number of layers, the number of nodes used in each layer, and the like. It may be stored as a parameter of the learning model 40 . Alternatively, for example, the parameter storage unit 41 may store, as parameters of the machine learning model 40, a calculation formula for obtaining a physical property value from a plurality of condition item data and coefficients of the calculation formula.

[2-3-2. Learning department]
The learning unit 42 updates the parameters of the machine learning model 40 stored in the parameter storage unit 41 by performing machine learning using a plurality of learning data stored in the learning data storage unit 20 . That is, the parameters of the machine learning model 40 stored in the parameter storage unit 41 are learned (updated) using a plurality of pieces of learning data. The learning unit 42 may be implemented mainly by the processor 11 of the physical property prediction device 10, or may be implemented by a processor of another information processing device connected to the physical property prediction device 10 via a network.

FIG. 5 is a diagram showing an example of machine learning performed by the learning unit 42. As shown in FIG. In this embodiment, as shown in FIG. 5, the machine learning model 40 is composed of a plurality of machine learning models learned for each of a plurality of attributes (for example, attributes 1 to 5 shown in FIG. 3). . In other words, the parameter storage unit 41 stores machine learning model parameters for each of a plurality of attributes.

As shown in FIG. 5, the learning unit 42 selects the condition item data (raw material 1 to raw material P shown in FIG. 3) to which attribute 1 is assigned among the plurality of condition item data included in the learning data. The machine learning model belonging to attribute 1 is trained using That is, among the parameters stored in the parameter storage unit 41, the parameters of the machine learning model to which attribute 1 is assigned are updated by the condition item data classified as attribute 1 in the learning data. Similarly, machine learning models belonging to attributes 2 to 5 are also learned by condition item data to which the corresponding attributes are assigned in the learning data. That is, the parameters of the machine learning model to which attributes 2 to 5 are assigned are also updated by the condition item data classified into the corresponding attributes in the learning data.

[2-3-3. Predictor]
When given input data (see FIG. 4) is input to the machine learning model 40, the prediction unit 43 outputs predicted physical property data for the input data. “Predicted physical property data” is data indicating predictions of properties of substances generated and measured under conditions indicated by input data (more specifically, a plurality of condition item data included in input data). In this embodiment, the "predicted physical property value" shown in FIG. 7 corresponds to predicted physical property data.

FIG. 6 is a diagram showing an example of prediction using a machine learning model performed by the prediction unit 43. As shown in FIG. As described above, in this embodiment, the machine learning model 40 is composed of a plurality of machine learning models learned for each of a plurality of attributes (for example, attribute 1 to attribute 5). Therefore, the prediction unit 43 classifies a plurality of condition item data included in the input data acquired by the input data acquisition unit 30 into a plurality of attributes (attribute 1 to attribute 5). Then, by inputting the condition item data of the input data classified for each attribute into the machine learning model of the corresponding attribute, the predicted physical property data for the input data is calculated.

FIG. 7 is a diagram showing an example of prediction results for input data. As shown in FIG. 7, the prediction unit 43 calculates a predicted physical property value, which is a numerical value indicating the property of a substance, as predicted physical property data for each of a plurality of input data. In the example shown in FIG. 7, the predicted physical property value of the input data identified by the ID "1" is "1.2", and the predicted physical property value of the input data identified by the ID "2" is "1. 2”. Note that the predicted physical property data is not limited to numerical values such as predicted physical property values, and may be data that can identify predicted physical properties. For example, the predicted physical property data may be represented by a sequence of numbers, symbols, characters, or the like.

[2-4. Similarity determination unit, predicted physical property data output unit]
The similarity determination unit 50 determines similarity (such as degree of similarity or degree of similarity) between the learning data and the input data. The predicted physical property data output unit 60 outputs predicted physical property data predicted by the prediction unit 43 of the machine learning model 40, and information based on the similarity determined by the similarity determination unit 50 as the accuracy of the predicted physical property data. to output The predicted physical property data output unit 60 outputs the predicted physical property value and the accuracy of the predicted physical property value to a display device such as the display unit 14, as shown in FIG. 7, for example. Note that the similarity determination unit 50 and the predicted physical property data output unit 60 may be realized mainly by the processor 11 of the physical property prediction device 10, or may be implemented by a processor of another information processing device connected to the physical property prediction device 10 via a network. and the like.

More specifically, the similarity determination unit 50 determines the similarity between each of the plurality of learning data and the input data. The similarity determination unit 50 determines, for example, each of the plurality of learning data identified by IDs "1", "2", . . . "N" in FIG. Similarity with input data is determined. Then, the predicted physical property data output unit 60 outputs the predicted physical property data, and outputs information based on any of the plurality of similarities determined by the similarity determination unit 50 as the accuracy of the predicted physical property data. For example, the predicted physical property data output unit 60 outputs predicted physical property data predicted by the prediction unit 43 of the machine learning model 40 for the input data identified by ID "1" in FIG. In addition, regarding the similarity between the input data and each of the plurality of learning data identified by IDs "1", "2", ..., "N" in FIG. outputs information based on any of the similarities between

In this embodiment, the similarity determination unit 50 determines similarity between the condition item data included in the learning data and the condition item data corresponding to the condition item data of the learning data included in the input data. Based on the similarity of the condition item data, the similarity between each of the plurality of learning data and the input data is determined.

In the examples shown in FIGS. 3 and 4, both the learning data and the input data include values of multiple items such as raw material 1 and raw material 2 as multiple condition items. The similarity determination unit 50 determines the similarity between the value of the learning data and the value of the input data for each of a plurality of items. The similarity determination unit 50 determines if the difference between the value of the learning data and the value of the input data for a certain item (raw material 1, etc.) is within a predetermined numerical range (in other words, if the absolute value of the difference is equal to or less than the threshold ), the item is determined to be similar, and if the difference is not within a predetermined numerical range (in other words, the absolute value of the difference exceeds the threshold), the item is determined to be dissimilar. In addition, the similarity determination unit 50 may determine the similarity of each of a plurality of items in three stages of matching, similarity, and non-matching. You may calculate the score according to.

The predicted physical property data output unit 60 outputs information based on the similarity determined for one of the plurality of learning data most similar to the input data as the accuracy of the predicted physical property data predicted for the input data. The predicted physical property data output unit 60 outputs, for example, information based on similarity to the learning data having the largest number of items determined to be similar among the plurality of learning data. In addition, for example, when the similarity determination unit 50 calculates a score according to the degree of similarity for a plurality of items, the predicted physical property data output unit 60 determines the similarity with the learning data having the highest calculated total score. Information based on gender may be output.

The plurality of condition item data included in the input data identified by ID "1" in FIG. 4 are compared with the plurality of condition item data included in the learning data identified by ID "1" in FIG. , raw material 1, raw material 2, raw material P, raw material P+1, raw material Q, raw material Q+1, raw material R, process condition 1, process condition S, measurement condition 1 (items other than measurement condition T) are consistent, and other The number of matching (similar) condition item data is the largest compared to the learning data. Therefore, the predicted physical property data output unit 60 outputs information based on similarity to the learning data identified by ID "1" as the learning amount data most similar to the input data identified by ID "1". to output

The predicted physical property data output unit 60 may output information based on the number of condition item data determined to be similar by the similarity determination unit 50 as the accuracy of the predicted physical property data. In this case, the predicted physical property data output unit 60, for example, when the number of condition item data determined to be similar is equal to or greater than a predetermined number, outputs “O (circle)”, and the number of condition item data determined to be similar is less than a predetermined number, output "x (x)". Not limited to this, the predicted physical property data output unit 60 may output the number of condition item data determined to be similar, or output the total score calculated for each of the plurality of condition item data. good too.

FIG. 8 is a diagram showing an example of a similarity determination method between learning data and input data. As described above, one of a plurality of attributes (attribute 1 to attribute 5) is assigned to the condition item data in each of the learning data and the input data. In this case, the similarity determination unit 50 determines the condition item data included in the learning data and the corresponding condition item data (learning data It is also possible to judge the similarity between the condition item data and the condition item data whose item name corresponds (matches), and based on that similarity, judge the similarity between the learning data for each attribute and the input data. good.

In addition, the predicted physical property data output unit 60 outputs information based on the number of condition item data determined to be similar by the similarity determination unit 50 and the total score value for each of a plurality of attributes (attribute 1 to attribute 5). may In the example shown in FIG. 7, the predicted physical property data output unit 60 outputs information such as "○, ○, ○, ○, ×" as the accuracy of the predicted physical property value for the input data specified by the ID "1". output. This is because the number of condition item data determined to be similar for each of attribute 1 to attribute 4 (or the total value of scores) is equal to or greater than the threshold defined for each corresponding attribute, and It indicates that the number of condition item data (or the total score value) determined to be true is less than the threshold defined by attribute 5. That is, attribute 1 to attribute 4 are all similar, but attribute 5 is not similar. In addition, the predicted physical property data output unit 60 outputs information such as "O, O, X, X, X" as the accuracy of the predicted physical property value for the input data specified by ID "2". This is because the number of condition item data (or the total score) determined to be similar for each of attribute 1 and attribute 2 is equal to or greater than the threshold for each corresponding attribute, and each of attribute 3 to attribute 5 indicates that the number of condition item data (or the total value of scores) determined to be similar in is less than the threshold of each corresponding attribute.

FIG. 9 is a diagram showing another example of prediction results for input data. As shown in FIG. 9, the predicted physical property data output unit 60 predicts the number of attributes for which the number of condition item data determined to be similar for each of a plurality of attributes (attribute 1 to attribute 5) is equal to or greater than a threshold. It may be output as the accuracy of the physical property value. For example, in the input data identified by ID "1", since the number of condition item data (or the total value of scores) of the four attributes of attribute 1 to attribute 4 is equal to or greater than the threshold value, the input data "4" is output as the accuracy of In the input data specified by ID "2", the number of condition item data in two attributes of attribute 1 to attribute 2 is equal to or greater than the threshold value, so "2" is output as the accuracy of the input data. there is

The certainty of physical property prediction based on multiple condition item data included in input data differs for each input data (that is, for each prediction). In this regard, according to the similarity determination unit 50 and the predicted physical property data output unit 60, the similarity between the learning data and the input data is output as the accuracy of the predicted physical property value. You can know the certainty.

[2-5. Item name output section]
The item name output unit 70 outputs the condition item data included in the learning data and the corresponding condition item data included in the input data (condition item data whose item name corresponds (matches) with the condition item data of the learning data). and are not similar, the item name of the condition item data is output. The item name output unit 70 may be realized mainly by the processor 11 of the physical property prediction device 10, or may be realized by a processor of another information processing device connected to the physical property prediction device 10 via a network. .

As shown in FIG. 9, the item name output unit 70 outputs the item name of the condition item data determined to be dissimilar as a cause of obstructing the similarity between the input data and the learning data. For example, the input data shown in FIG. 3 whose ID is "1" and the learning data shown in FIG. 4 whose ID is most similar to the input data specified by "1" means that the absolute value of the difference between the values of the condition item data with the item name "measurement condition T" is "3", which is greater than a predetermined threshold value (for example, "1"). Therefore, the item name output unit 70 outputs the item name of "measurement condition T" for the input data specified by ID "1". Further, the input data identified by ID "2" and the learning data most similar to the input data identified by ID "1" are "raw material Q", "process condition 1", "measurement The value difference in the item name of condition T" is not within the predetermined numerical range. Therefore, the item name output unit 70 outputs the item names of "raw material Q", "process condition 1", and "measurement condition T" for the input data specified by ID "2".

As described above, the item name output unit 70 outputs the item name of the condition item data determined to be dissimilar as a cause of obstructing the similarity between the input data and the learning data. It is possible to know which item is not similar to the learning data, which is known data. Based on this information, the user can make a decision as to whether or not to make a prototype of the substance, or use it to create input data (redesign).

[3. flowchart]
FIG. 10 is a diagram showing an example of the flow of prediction processing performed by the physical property prediction device 10. As shown in FIG. First, the input data acquisition unit 30 acquires input data including a plurality of condition item data (step S101). The number of input data acquired in step S101 may be plural as shown in FIG. and may be one.

Next, the prediction unit 43 generates a machine learning model for each attribute (more specifically, based on the learning data stored in the learning data storage unit 20, the machine learned for each attribute 1 to attribute 5). parameters of the learning model), the predicted physical property value of the input data is calculated (step S102). At this time, the prediction unit 43 classifies a plurality of condition item data included in the input data acquired in step S101 into a plurality of attributes (attributes 1 to 5), and corresponds the condition item data classified for each attribute. May be input into machine learning models for attributes. Thereby, the predicted physical property value for the input data is calculated. When a plurality of input data are obtained in step S101, predicted physical property values are calculated for each of the plurality of input data in step S102.

Next, the similarity determination unit 50 determines similarity for each attribute between the input data acquired in step S101 and the learning data stored in the learning data storage unit 20 (step S103). The similarity determination unit 50 determines the similarity between the input data and each of the learning data of the plurality of learning data stored in the learning data storage unit 20 . More specifically, condition item data included in the learning data, condition item data included in the corresponding input data (condition item data whose item name corresponds (matches) with the condition item data of the learning data), , and based on the similarity, the similarity between the input data and the learning data is determined.

The similarity determination unit 50, for example, determines the similarity of the condition item data included in the input data and the learning data for each of a plurality of attributes (attributes 1 to 5), and determines the similarity between the input data and the learning data based on the similarity. The similarity (degree of similarity, degree of similarity, etc.) for each attribute with the learning data is determined.

Next, the predicted physical property data output unit 60 outputs information based on the predicted physical property value calculated in step S102 and the degree of similarity determined in step S103 (step S104), and ends the process. The predicted physical property data output unit 60 outputs information based on the similarity determined for the learning data most similar to the input data among the plurality of learning data stored in the learning data storage unit 20 as a predicted physical property value. is output as the accuracy of The predicted physical property data output unit 60 outputs, for example, information indicating similarities determined for each of a plurality of attributes (attributes 1 to 5). The predicted physical property data output unit 60 may output, for example, characters including "o" and "x" shown in FIG. You can output a number.

Calculation of predicted physical property values is performed based on machine learning models machine-learned for each attribute (step S102). For this reason, the similarity for each attribute between the input data and the learning data is determined (step S103), and the determined similarity for each attribute is output as the probability (step S104). can increase the likelihood of the probability of indicating

In step S104, the item name output unit 70 outputs the condition item data included in the learning data and the corresponding condition item data included in the input data. condition item data), the item name of the condition item data may be output. For example, as shown in FIG. 9, the item name output unit 70 may output the item name of the condition item data determined to be dissimilar as a cause of obstructing the similarity between the input data and the learning data. good. By doing so, the user can know in which item the input data is not similar to the learning data, which is known data.

[4. summary]
As described above, in the present embodiment, the similarity determination unit 50 determines the similarity between each of the plurality of learning data and the input data. The predicted physical property data output unit 60 outputs predicted physical property data indicating the properties of the substance generated under the production conditions indicated by the input data, and the similarity determined by the similarity determination unit 50 is used as the accuracy of the predicted physical property data. Output information based on either By doing so, the similarity between the learning data and the input data is output as the accuracy of the predicted physical property value, so that the user can know the certainty of the predicted physical property value for each input data (that is, for each prediction). be able to. The user can make use of it in determining which prediction to rely on when fabricating an actual material.

[5. Modification]
The present invention is not limited to the above embodiments.

In the embodiment, the similarity determination unit 50 compares the condition item data included in the learning data and the condition item data included in the input data (the condition item data and item names of the learning data correspond (match)) for each attribute. An example has been described in which the similarity with the condition item data) is determined, and the similarity is determined based on the similarity. The similarity determination unit 50 may determine the similarity between the learning data and the input data by various methods.

For example, the similarity determination unit 50 generates attribute data based on condition item data to which each attribute is assigned for each of a plurality of attributes with respect to the learning data and the input data, and compares the attribute data related to the learning data and the input data. The similarity between the data and attribute data may be determined, and the similarity between the learning data and the input data may be determined based on the similarity. Attribute data is data provided for each attribute (for example, for each of a plurality of attributes 1 to 5) and includes at least one element. The elements included in the attribute data of each attribute may differ from or match the contents of the condition item data of the corresponding attribute.

11A to 11C are diagrams showing an example of attribute data generated for each of attribute 1 to attribute 3 (attributes assigned to raw material data). The example shown in FIG. 11A shows attribute data generated for attribute 1 and includes a plurality of elements. In the example shown in FIG. 11A, "element 1" is the same element as "raw material 1", "element 2" is the same element as "raw material 2", and "element X" is the same element as "raw material P". It is included. That is, the multiple elements included in the attribute data of attribute 1 match the multiple condition item data to which attribute 1 shown in FIGS. 3 and 4 is assigned.

The example shown in FIG. 11B shows the attribute data generated for attribute 2, which includes "element X+1" which is the same element as "raw material 1", "element X+2" which is the same element as "raw material P+1", "Element Y" which is the same element as "raw material Q" is included. Here, "ingredient 1" is included not only in the attribute 2 element but also in the attribute 1 element. In this way, the similarity determination unit 50 may generate two pieces of attribute data including specific condition item data.

In the example shown in FIG. 11C, attribute data generated for attribute 3 is shown. A certain "element Y+2" and "element Z" which is the same element as "raw material R" are included. As shown in FIG. 4, in the input data specified by ID "1", the value of raw material 1 is "44" and the value of raw material 2 is "10". , "54", which is the sum of "44" and "10", is set to "element Y+1" which is an element combining "raw material 1" and "raw material 2" related to the input data. Similarly, the similarity determination unit 50 also sets "54" to the "element Y+1" related to the learning data.

In this manner, the similarity determination unit 50 may generate attribute data related to the learning data and attribute data related to the input data for the attributes 1 to 3. FIG. Then, the similarity between the attribute data related to the learning data and the attribute data related to the input data (more specifically, the similarity between elements having the same item name) is determined, and learning is performed based on the similarity. The similarity between the data for use and the input data may be determined. The similarity determination unit 50 generates attribute data for attribute 4 (attribute assigned to process condition data) and attribute 5 (attribute assigned to measurement condition), which are other attributes, in the same manner as attributes 1 to 3. Alternatively, similarity of condition item data to which attribute 4 and attribute 5 are assigned may be determined as described in the embodiment.

The similarity determination unit 50 determines similarity between a plurality of condition item data associated with each of the plurality of learning data stored in the learning data storage unit 20 and a plurality of condition item data associated with the input data. Therefore, when the number of multiple condition item data included in each of the learning data and the input data is enormous, the processing time becomes long. In this regard, the similarity determination unit 50 generates attribute data including an element combining a plurality of condition item data as shown in FIG. and the number of condition item data included in the input data. By judging the similarity of the attribute data for each of the learning data and the input data, the processing time is shortened compared to directly comparing the plurality of condition item data included in the learning data and the input data. be able to.

The predicted physical property data output unit 60 outputs the similarity between the learning data determined in this way and the input data as the accuracy of the predicted physical property value. This also allows the user to know the likelihood of the predicted physical property value for each piece of input data.

10 physical property prediction device 11 processor 12 storage unit 13 communication unit 14 display unit 15 operation unit 20 learning data storage unit 30 input data acquisition unit 40 machine learning model 41 parameter storage unit 42 learning unit , 43 prediction unit, 50 similarity determination unit, 60 predicted physical property data output unit, 70 item name output unit.

Claims (8)

learning condition data storage means for storing a plurality of learning condition data respectively indicating substance production conditions;
A parameter of a machine learning model that is learned using the plurality of learning condition data and that outputs predicted physical property data indicating the properties of a substance produced under the production conditions indicated by the input condition data when given input condition data is input. a parameter storage means for storing
similarity determination means for determining similarity between each of the plurality of learning condition data and the input condition data;
output means for outputting the predicted physical property data and outputting information based on any of the similarities as the accuracy of the predicted physical property data;
A physical property prediction device comprising: In the physical property prediction device according to claim 1,
The output means outputs, as the accuracy, information based on the similarity determined for one of the plurality of learning condition data most similar to the input condition data.
A physical property prediction device characterized by: In the physical property prediction device according to claim 1 or 2,
both the learning condition data and the input condition data include a plurality of condition item data;
The similarity determination means determines similarity between the condition item data included in the learning condition data and the corresponding condition item data included in the input condition data, and based on the similarity, determining the similarity;
A physical property prediction device characterized by: In the physical property prediction device according to claim 3,
When it is determined that the condition item data included in the learning condition data and the corresponding condition item data included in the input condition data are not similar, the item name of the condition item data is output. further comprising an item name output means;
A physical property prediction device characterized by: In the physical property prediction device according to claim 3 or 4,
Each condition item data is assigned one of a plurality of attributes,
The similarity determination means determines similarity between the condition item data included in the learning condition data and the corresponding condition item data included in the input condition data for each attribute. determining the similarity based on no;
A physical property prediction device characterized by: In the physical property prediction device according to claim 1 or 2,
both the learning condition data and the input condition data include a plurality of condition item data;
Each condition item data is assigned one of a plurality of attributes,
The similarity determination means generates attribute data based on the condition item data to which the attribute is assigned for each attribute of the learning condition data and the input condition data, and generates attribute data related to the learning condition data. and attribute data related to the input condition data, and determining the similarity based on the similarity;
A physical property prediction device characterized by: In the physical property prediction device according to any one of claims 1 to 6,
The learning condition data and the input condition data are further data indicating measurement conditions of substances,
The predicted physical property data is further data indicating properties of the substance measured under the measurement conditions indicated by the input condition data.
A physical property prediction device characterized by: learning condition data storage means for storing a plurality of learning condition data each indicating a substance production condition;
A parameter of a machine learning model that is learned using the plurality of learning condition data and that outputs predicted physical property data indicating the properties of a substance produced under the production conditions indicated by the input condition data when given input condition data is input. parameter storage means for storing
Similarity determining means for determining similarity between each of the plurality of learning condition data and the input condition data; and outputting the predicted physical property data and outputting the similarity as accuracy of the predicted physical property data. output means to
A program that allows a computer to function as a

Images