JP7330712B2 - Material property prediction device and material property prediction method - Google Patents

Description

The present invention relates to a material property prediction device and a material property prediction method, and in particular, predicts material properties based on experimental data at the time of material development, and evaluates using human knowledge when selecting an appropriate material. It relates to a material property prediction device and a material property prediction method suitable for performing.

With the development of statistical processing technology for data analysis, there is a growing demand for data analysis in materials science as well. In particular, in the field of material science, a method called screening is known, in which candidates for the next experiment are selected based on known data in order to efficiently develop new materials.

As a screening method, data from various experiments are input into an information system and subjected to machine learning to construct a model for predicting experimental results. A well-known method is to find a function that returns a characteristic by regression analysis. In the following description, a model representing material properties represented by such a function is called a material property prediction model.

A system for analyzing data on compounds by a computer is described, for example, in Patent Document 1. In Patent Document 1, chemical data is visualized by mapping it on a map window (Fig. 6) in a two-dimensional space, and the user edits the visualized data to obtain a two-dimensional compression result. A screening method is disclosed in which the image is adjusted so that it is easy to see.

Also, a technique for high-speed screening for drug discovery is described in Patent Document 2, for example. In Patent Document 2, in order to derive a rule that connects structural features and biological activities from relatively small screening data, parameters are partially used in a subspace, which is a space that has good material properties for similar data. A screening method for preferentially testing candidates with a large number of .

Japanese Patent Publication No. 2001-503546 Japanese Patent Publication No. 2005-506511

In material development, by increasing the accuracy of the material property prediction model, it will be possible to more accurately determine the promise of new material candidates, and by omitting unnecessary experiments, it will be possible to develop materials efficiently. Be expected.

However, in practice, it is not easy to improve the accuracy of material property prediction. One of the major reasons for this is the large amount of data required to make highly accurate predictions. One of the causes is the large number of material parameters. Without knowing which parameters contribute to the material properties, all parameters are used in the prediction. In such a case, a phenomenon generally called over-learning occurs in machine learning, resulting in a decrease in accuracy. It is conceivable to increase the number of experiments to compensate for this, but the number of data required for prediction increases exponentially with the number of material parameters, so a considerable number of experiments must be performed. Then, there may be a situation where it would have been quicker to direct the effort of the experiment to the search for new materials.

Therefore, instead of relying only on prediction, there is a technique that facilitates screening by appropriately projecting and compressing data with many parameters into a two-dimensional space and visualizing it.

The analysis system described in Patent Document 1 visualizes chemical data in a mapping window in a two-dimensional space, and the user edits the visualized data, resulting in a two-dimensional compression result. It adjusts and screens so that it may become easy to see. This method facilitates visual screening. However, in addition to this, it is necessary for the user to consider everything in order to reflect the result of material property prediction, which requires labor.

In addition, the analysis method described in Patent Document 2 is a method of preferentially experimenting candidates with many similar data and good material properties in a subspace that is a space that partially uses parameters. there were. According to this method, screening can be performed even when there are many parameters, but in situations where there is not much data and there are too many parameters, it may be judged as if there is a correlation when there is actually no correlation. This is a phenomenon similar to the above-mentioned over-learning, and the point that it is not easy to improve accuracy without selecting parameters is still unsolved.

The purpose of the present invention is to predict the characteristics of materials based on experimental data at the time of material development, and when selecting appropriate materials, in the prediction value evaluation for screening of experimental plans, human knowledge is utilized. An object of the present invention is to provide a material property prediction device and a material property prediction method that can be evaluated.

The material property prediction device of the present invention is preferably a material property prediction device for predicting and displaying material properties, comprising: a material data table storing material properties and material parameters associated with the materials; A material data visualization table that stores the result of dimensional compression as two-dimensional coordinates, and a parameter importance table that stores the parameter importance associated with each material parameter. A dimensional compression unit without teacher data that stores the coordinates in the material data table, and a 2D visualization unit that two-dimensionally displays the two-dimensional coordinates stored in the material data visualization table as material data points and accepts position changes of the material data points. and, based on the received repositioned material data point information, perform a regression analysis to update the parameter importance stored in the parameter importance table and compare the repositioned material data points with the unrepositioned material data points. A dimensional compression part with semi-supervised data that dimensionally compresses the material parameters of the material data points and updates them as two-dimensional coordinates stored in the material data visualization table, and a material parameter similarity that weights the parameter importance to each material parameter. and a predictive evaluator for predicting material properties by performing regression analysis based on the degree.

According to the present invention, when predicting the characteristics of materials based on experimental data during material development and selecting appropriate materials, human knowledge is utilized in predictive value evaluation for screening experimental plans. It is possible to provide a material property prediction device and a material property prediction method that can be evaluated.

2 is a functional configuration diagram of the material property prediction device of Embodiment 1. FIG. 1 is a hardware/software configuration diagram of a material property prediction device; FIG. It is a figure which shows an example of a material data table. It is a figure which shows an example of a material data visualization table. It is a figure which shows an example of a material and coordinate mapping table. It is a figure which shows an example of a parameter importance table. It is a flowchart which shows the outline|summary of the process by a material property prediction apparatus. 4 is a flowchart showing an example of material data visualization processing according to the first embodiment; 9 is a flowchart showing an example of parameter importance editing processing; FIG. 11 is a flowchart showing an example of dimension compression processing with parameter importance; FIG. 6 is a flowchart showing an example of material property prediction presentation processing; 5 is a flow chart showing an example of a material data 2D display screen of Embodiment 1. FIG. 4 is a flow chart showing an example of a material property prediction presentation screen; 2 is a functional configuration diagram of a material property prediction device of Embodiment 2. FIG. It is a figure which shows an example of a material visualization background table. 10 is a flowchart showing an example of material data visualization processing according to the second embodiment; FIG. 10 is a diagram showing an example of a material data 2D display screen according to Embodiment 2;

Hereinafter, each embodiment according to the present invention will be described with reference to FIGS. 1 to 17. FIG.

[Embodiment 1]
A first embodiment according to the present invention will be described below with reference to FIGS. 1 to 13. FIG.

First, the configuration of the material property prediction device of Embodiment 1 will be described with reference to FIGS. 1 and 2. FIG.
The material property prediction device 101 is a device that presents necessary information for predicting material properties from information on materials in accordance with a user's operation.

As shown in FIG. 1 , the material property prediction device 101 has, as a functional configuration, a prediction evaluation unit 112, a dimension compression unit 113 without teacher data, a 2D (Dimension) visualization unit 115, a dimension compression unit with semi-supervised data 116, a material It consists of the property prediction presentation unit 117 .

The prediction evaluation unit 112 is a functional unit that weights the parameters of the material and calculates the prediction accuracy. The non-teacher data dimension compression unit 113 is a functional unit that gives multidimensional material parameters and converts them into two-dimensional coordinates. Here, "without training data" is in contrast to "with semi-teaching data", which will appear later. , means that the dimension is compressed as it is. The 2D visualization unit 115 is a functional unit that visualizes two-dimensional coordinates obtained by converting material parameter information. A dimension compression unit with semi-supervised data 116 is a functional unit that dimension-compresses the material parameters again based on the user editing result of the material parameters and the two-dimensional coordinate data resulting from the dimension compression, and calculates the dimensional coordinates. is. "With semi-supervised data" means that the user's user interface is used to edit the information on the material data, and the data generated thereby is used to newly perform dimensional compression. The material property prediction presentation unit 117 is a functional unit that finally presents the prediction of material properties to the user.

The material property prediction device 101 also holds a material data table 240, a material data visualization table 250, a parameter importance level table 260, and a material/coordinate mapping table 270 as data.

The material data table 240 is a table that stores information on material properties and parameters given to materials. A material data visualization table 250 is a table that stores information-converted two-dimensional coordinates of material data and information related thereto. The parameter importance table 260 is a table that stores the parameter importance (details will be described later) of each parameter used in the material prediction calculation process. The material/coordinate mapping table 270 is a table that stores the correspondence between the material and the two-dimensional coordinates when the user edits the two-dimensional display point of the material displayed on the screen (details will be described later).

Next, the hardware/software configuration of the material property prediction device will be described with reference to FIG.

The material property prediction device 101 can be realized by a general information processing device such as a personal computer or workstation. Material property prediction device 101, for example, as shown in FIG. are connected by a bus.

The CPU 201 controls each part of the material property prediction device 101, loads necessary programs into the main memory 202, and executes them.

The main memory 202 is normally composed of a volatile memory such as a RAM, and stores programs to be executed by the CPU 201 and data to be referred to.

A network I/F 206 is an interface for communicating with other devices via a network connected to the material property prediction device 101 .

A display I/F 203 is an interface for connecting a display device 210 such as an LCD (Liquid Crystal Display).

The input/output I/F 204 is an interface for connecting input/output devices. In the example of FIG. 2, a keyboard 221 and a mouse 222 as a pointing device are connected.

Auxiliary storage I/F 205 is an interface for connecting auxiliary storage devices such as HDD (Hard Disk Drive) 230 and SSD (Solid State Drive).

The HDD 230 has a large storage capacity and stores programs for executing this embodiment. A prediction evaluation program 231, a dimension compression program without teacher data 232, a 2D visualization program 233, a dimension compression program with semi-teacher data 234, and a material property prediction presentation program 235 are installed in the material property prediction device 101. FIG.

The prediction evaluation program 231, the dimension compression program without teacher data 232, the 2D visualization program 233, the dimension compression program with semi-supervised data 234, and the material property prediction presentation program 235 are the prediction evaluation unit 112, the dimension compression unit without teacher data 113, and the 2D visualization program 233, respectively. It is a program that executes the functions of a visualization unit 115 , a dimension compression unit with semi-supervised data 116 , and a material property prediction presentation unit 117 .

The HDD 230 also stores a material data table 240 , a material data visualization table 250 , a parameter importance level table 260 and a material/coordinate mapping table 270 .

Next, the data structure handled by the material property prediction device will be described with reference to FIGS. 3 to 6. FIG.

The material data table 240 is a table that stores information about materials, and as shown in FIG. 3, it is a table composed of records consisting of items of material ID 501, material name 502, material properties 503, and material parameters 504. .

The material ID 501 stores an identifier that uniquely identifies the material. The material name 502 stores the name of the material. The material property 503 stores a numerical value representing the property of the material that has been found by experiment or the like. The material properties 503 store numerical values that have been found through experiments, but may be blank if no experiments have been conducted on the material. The material parameter 504 stores parameters for calculating material characteristics such as material structure information and manufacturing conditions. The material parameters 504 are stored in a format such as (parameter 1, value 1), (parameter 2, value 2), . The material parameters 504 may include all kinds of information indicating the structure and production process of the material, such as chemical structural formulas.

The material data visualization table 250 is a table that stores work data related to materials for visualizing material data. As shown in FIG. The table is composed of records having items of accuracy evaluation value 605 , X coordinate 606 , Y coordinate 607 , and material property prediction value 608 .

A material ID 601, a material property 602, and a material parameter 603 store duplicated values of the material ID 501, material property 503, and material parameter 504 in the material data table 240, respectively. The edit presence/absence flag 604 stores a flag indicating whether or not the material indicated by the material ID 501 has been edited by the user using a material data 2D display screen (described later). The initial value of this edit presence/absence flag 604 is a FALSE value. The prediction accuracy evaluation value 605 stores a prediction accuracy evaluation value obtained by cross-validation of the prediction evaluation unit 112 (details will be described later). It should be noted that no prediction is made for materials whose material property 602 is blank, and the prediction accuracy evaluation value 605 is stored as blank. For the X coordinate 606 and Y coordinate 607, values obtained when the values of the material parameter 504 are dimensionally compressed (details will be described later) and made two-dimensional are possible. If the material property 602 exists, it is stored in the material property prediction value 608; otherwise, the predicted value of the material property by the prediction evaluation unit 112 is stored.

The material/coordinate mapping table 270 is a table that stores the result of editing by the user using a material data 2D display screen (described later). As shown in FIG. This table is composed of records consisting of 902 items. In the material/coordinate mapping table 270, when the user edits data on the material data 2D display screen, the edited X coordinate 901 and Y coordinate 902 corresponding to the edited data are described in the material ID 601. It is recorded in a form associated with the ID.

The parameter importance level table 260 is a table that stores the parameter importance level for each parameter, and is a table composed of records consisting of items of parameter ID 1101 and parameter importance level 1102 .

The parameter importance level 1102 stores the parameter importance level value for the parameter ID stored in the parameter ID 1101 (the parameter ID described in the material parameter 504 in FIG. 3). It should be noted that the parameter ID shall be a unique identifier for all materials. Although the details of the parameter importance will be described later, when the X coordinate and the Y coordinate are obtained by newly compressing the dimension of the material parameter in consideration of the data edited by the user from the material data 2D display screen, It is evaluated as a kind of weighting for each material parameter in the kernel function when constructing the prediction function by regression analysis.

Next, the processing of the material property prediction device will be described with reference to FIGS. 7 to 13. FIG.

First, with reference to FIG. 7, an outline of processing when a user makes a material prediction using a material property prediction device will be described.
The material property prediction device 101 executes material data visualization processing and displays a material data 2D display screen 312 to the user ((I) initial map presentation). The material data 2D display screen, which will be described in detail later, is a screen that displays the result of compressing the material parameter information shown in the material data visualization table 250 into a two-dimensional space and information related thereto. The operator performs data manipulation on the display screen from the material data 2D display screen 312, and along with this, the material property prediction device 101 performs parameter importance editing processing ((II) map editing/updating). As a result, the parameter importance is corrected in accordance with the knowledge of the user. This processing updates the parameter importance. The user then repeatedly executes this data manipulation until he is satisfied, and the record is accumulated as the material-coordinate mapping table 270, and is finally reflected in the material data visualization table.

Finally, when the map editing/updating process is finished, the material property prediction device 101 executes the material property prediction presentation processing 331 and presents the material property prediction display screen to the user ((III) prediction result viewing). As a result, the user can grasp the material properties in which the data manipulation in the (II) map editing/updating phase is reflected.

Next, material data visualization processing will be described with reference to FIG.
In the material data visualization process, first, the prediction evaluation unit 112 calculates a prediction accuracy evaluation value (details will be described later) by cross-validation (details will be described later) based on the material data table 240 (S100).

A detailed description of this process is as follows. The prediction evaluation unit 112 acquires only the material data visualization table 250 in which the material property 602 is described, and calculates the prediction accuracy evaluation value of the material property based on cross-validation. For those in which the material properties 602 are not described, a material property prediction function is constructed using data describing the material properties 602 as learning data (details will be described later), and the predicted values of the material properties are calculated. . The results are stored in the prediction accuracy evaluation value 605 and material property prediction value 608 of the material data visualization table 250 .

The prediction evaluation unit 112 constructs a function (hereinafter referred to as a “material property prediction function”) that returns a prediction value of the material property when the value of the material parameter described in the material parameter 504 of the material data table 240 is given as an argument. have a function. This function predicts material properties based on the degree of similarity between two pieces of data, and can be realized by a known kernel regression analysis method, support vector regression, or the like. The prediction evaluation unit 112 is characterized in that it can be used by weighting each variable of the material parameter based on the importance of the parameter in calculating the degree of similarity between data (details will be described later). Since the parameter importance corresponds to the initial value, it may be determined by any method, for example, it may be fixed at a predetermined value, or it may be determined randomly.

Here, cross-validation means that in statistics, sample data is divided, part of the sample data is used as learning data, the rest is used as evaluation data, and the learning data is first analyzed, This is a method of testing the analysis using the evaluation data to verify and confirm the validity of the analysis itself. In the cross-validation of this embodiment, a material property prediction function is constructed from the learning data of the material data, and the material parameter of the evaluation data is input to calculate the predicted value of the material property. Then, the difference between the material properties originally included in the evaluation data (hereinafter referred to as "true value") and the predicted value is obtained as an error. The smaller the absolute value of this error, the more accurate the reference prediction accuracy evaluation value is obtained. As a method for calculating the prediction accuracy evaluation value from this prediction value, any method may be used as long as the method is recognizable as accuracy by the user. For example, if a value obtained by dividing the error by the true value is used, the effect of increasing the error as the number of true values increases can be mitigated for evaluation. Alternatively, if the reciprocal of the absolute value of the error is used, it can be used as an intuitive index that the larger the value, the better the accuracy. Alternatively, by obtaining the logarithm of the ratio of the true value to the predicted value, we took into consideration the fact that it is difficult to improve further when the accuracy is already high (for example, to make the original 80% correct state 89% correct , it cannot be said that it is ten times easier to change a 99% correct state from the original to a 99.9% correct state).

Next, the non-teaching data dimension compression unit 113 generates X coordinates and Y coordinates of the material parameters by a dimension reduction technique for multivariates, and stores them in the X coordinates 606 and Y coordinates 607 of the material data visualization table 250 ( S101). Here, the dimensionality reduction method for multivariate means that when multivariate data is regarded as point data in a multidimensional space, the relationship between the point data is not damaged as much as possible (characteristics required for statistical analysis ), mapping as point data in a low-dimensional space, principal component analysis, subspace method, independent component analysis, non-negative matrix decomposition, singular value matrix decomposition, self-organization A map method, a t-SNE method, and the like are known.

Finally, based on the above results, the 2D visualization unit 115 generates display information for the material data 2D display screen 312 and presents it to the user through the display device 219 (S102).

Details of the material data 2D display screen 312 will now be described with reference to FIG.
The material data 2D display screen 312 consists of display elements such as a material data display frame 701, a material detailed information display frame 707, a color designation radio button 708, an update button 709, and a completion button 710, as shown in FIG.

A material data display frame 701 is an area for plotting point data (hereinafter referred to as “material data points”) mapped to X coordinates and Y coordinates by dimensionally compressing material parameters. Material data points are plotted in the material data display frame 701 so that the characteristics of the material data can be easily grasped. In the example shown in FIG. 12, blanks, ie, unmeasured material properties are indicated by white circles 702, black diamonds 703 for those with high prediction accuracy evaluation values 605, and gray diamonds 705 for those with high prediction accuracy evaluation values. , the lower ones are represented in different forms such as white diamonds 704 .

A material detailed information display frame 707 is an area for displaying detailed information of material data. When the user operates the cursor 706 using the mouse 222 and clicks a material data point in the material data display frame 701, detailed information about the material is displayed in the material detailed information display frame 707 at the lower left. The displayed information is created based on the information of the material parameter 603 of the material data visualization table 250, and the characteristics of the material are displayed in an easy-to-understand manner.

A color coding designation radio button 708 is a button that is set when the user switches the designation of color coding. Here, an example of switching based on the prediction accuracy and material properties of the material is shown.

The update button 709 is a button for specifying when the user's changes are reflected in the data, and the finish button 710 is a button for the user to finish the work and close the material data 2D display screen 312. .

Next, parameter importance editing processing will be described with reference to FIG.
In the parameter importance editing process, the material property prediction device first presents the material data 2D display screen 312 shown in FIG. 12 and accepts the user's data movement input (S200). The user can move the position of each point by operating the cursor 706 and dragging each point within the material data display frame 701 using a pointing device such as the mouse 220 .

As is clear from the characteristics of the dimension compression method described above, basically, it is expected that materials having relatively similar material parameters are placed close to each point in the material data display frame 701 . Also, if there is a tendency that material properties are similar when material parameters are similar, it is expected that the prediction evaluation unit 112 that makes predictions using the similarity of material parameters will have high prediction accuracy. Conversely, when there are points with low prediction accuracy evaluation values (white diamonds 704) shown in FIG. 12, the closeness of the material properties does not match the similarity of the material parameters. This leads to improper placement of the point within the material data display frame 701 .

The material property prediction method of the present invention focuses on the relationship between this prediction accuracy and the validity of the arrangement within the material data display frame 701 . That is, since increasing the prediction accuracy and increasing the validity of the placement within the material data display frame 701 are similar, by linking them according to the parameter importance, the placement within the material data display frame 701 It is possible to adjust the parameter importance so as to increase the prediction accuracy by a simple method of correcting .

In this embodiment, the user can make corrections by moving data with low prediction accuracy evaluation values to other points that are considered to be more similar in terms of material properties. After that, when the update button 709 is pressed to transfer the data (S201: YES), the material property prediction device 101 receives the transfer result (S202). At this time, the editing result is recorded in the material/coordinate mapping table 270 as a work table, and finally stored in the X coordinate 606 and Y coordinate 607 of the material data visualization table 250 . Then, the editing presence/absence flag 604 is changed from FALSE (false) to TRUE (true).

Next, the material property prediction device 101 executes dimension compression execution processing with semi-supervised data for re-executing dimension compression based on this data editing information (S203). Details of the dimension reduction execution process with semi-supervised data will be described later with reference to FIG.

Next, prediction evaluation unit 112 performs a process of evaluating prediction accuracy using parameter importance table 260 (S204).

Here, the method of regression analysis for constructing the dimension reduction function and the significance of the parameter importance in this embodiment will be described in detail. In this embodiment, a known Kernel Gaussian Process Regression (KGPR) method is used. In general, KGPR constructs a prediction function using a kernel function, which is a function that indicates the degree of similarity between two pieces of data. This weighted Euclidean distance means D({m _i }, {n _{i }} ) defined by the following (Equation 1) for two material parameters {m _i }, {n _i }.

D({m _i }, {n _i })=Σ _i γ _i (m _i −n _i ) ² (equation 1)

where γ _i is a parameter indicating the parameter importance of material parameters, and Σ _i means sum over all material parameters. In this embodiment, a Gaussian function represented by the following (Equation 2) using this weighted Euclidean distance is used as a kernel function.
k({m _i }, {n _i })=exp(−D({m _i }, {n _i })) (Formula 2)

In the well-known regression analysis of KGPR, kernel function values are obtained for all data used for learning, and prediction functions are constructed using the values. At the same time, it is possible to calculate the likelihood that indicates the degree of fit between the data used for learning and the prediction function. Here, γ _i of the kernel function can be determined by a known optimization method (for example, gradient descent method) so as to maximize the likelihood.

In the process of evaluating the prediction accuracy using the parameter importance table 260, the prediction evaluation unit 112 performs prediction using this KGPR and evaluates the prediction accuracy. That is, for the two material parameters {m _i } and {n _i }, expressed by (Equation 2) using D ({m _i }, {n _i }) defined in (Equation 1) above A prediction function of the material properties is obtained with the Gaussian function obtained as the kernel function.
At this time, the value of parameter importance 1102 of parameter importance table 260 is used for γ _i .

As will be described in detail later, in the dimension compression processing with semi-supervised data in S203, the editing result of the material data points on the material data 2D display screen of the user is reflected in the parameter importance level 1102 of the parameter importance level table 260. be. Therefore, this makes it possible to make a prediction that reflects the knowledge of the user. The parameter importance level 1102 of the parameter importance level table 260 may not be used as it is, but may be set so that it does not greatly differ from the value of the parameter importance level 1102 of the parameter importance level table 260 . For example, a method of optimizing the remaining parameters using the gradient descent method in the same way as when constructing the dimension reduction function is conceivable. In addition, a method of performing optimization with constraints so as to optimize within a range that does not greatly differ from the parameter importance 1102 of the parameter importance table 260, and a method of penalizing the degree of deviation from the parameter importance 1102 of the parameter importance table 260 However, any method may be used as long as it can reflect the parameter importance 1102 of the parameter importance table 260 .

Next, details of the dimension compression processing with semi-supervised data will be described with reference to FIG.
This is the process shown in S203 of the parameter importance editing process shown in FIG.
In S203 of the parameter importance editing process, first, a record in which the editing presence/absence flag 604 of the material data visualization table 250 is TRUE is obtained (S300).

Next, regression analysis is performed on the data of those records (S301). This regression analysis is a process of constructing a dimensional compression function g of the following (Equation 3) that gives an X coordinate 901 and a Y coordinate 902 after editing when material parameters are given as arguments.
(X, Y)=g({m _i }, {γ _i }) (Formula 3)
Here, (X, Y) are the values of the X coordinate 901 and the Y coordinate 902 after editing respectively, {m _i } is the value of the material parameter 603 of the material data visualization table 250, and {γ _i } is the material It is the value of the parameter importance level 1102 of the parameter importance level table 260 corresponding to the material parameter indicated by the parameter 603 . This method uses the same method as when constructing the material property prediction function of the prediction evaluation unit 112 described above, except that only the objective variable is different.

That is, let γ _i be the parameter importance corresponding to the material parameter _{m i} , and D({m _{i }} , { n _i }) using the Gaussian function represented by (Equation 2) as the kernel function to determine the prediction function of the material properties (prediction by KGPR).

When this method is applied to the present embodiment, it is possible to obtain a dimensional compression function that projects the material parameters to the edited X-coordinate 901 and Y-coordinate 902, and at the same time determine the parameter importance γ _i at that time.

When the dimension compression function and the parameter importance are obtained by the above procedure, from the material data for visualization in the material data visualization table 250, those with the edit flag 604 set to FALSE are acquired (S302), and those with the edit flag 604 set to TRUE are obtained. In addition, by inputting the material parameters into the dimensional compression function, new X coordinates 606 and Y coordinates 607 are obtained and set in the material data visualization table 250 (S303). At this time, if the editing flag 604 is TRUE, the X coordinate 901 and the Y coordinate 902 after editing may be used as they are to simplify the calculation.

The parameter importance obtained in the process of obtaining this dimension compression function is recorded in the parameter importance 1102 of the parameter importance table 260 (S304).

Next, details of the material property prediction presentation processing will be described with reference to FIG. 11 .
In the material property prediction presentation process, the material data (material ID 601, material property 602, material parameter 603, material property prediction value 608) is acquired from the material data visualization table 250 (S400), and from the parameter importance table 260, the parameter ID 1101 , the parameter importance 1102 is obtained (S401), and the result of predicting the material properties is displayed on the material property prediction presentation screen 332 based on the parameter importance (S402).

Here, the material property prediction presentation screen will be described with reference to FIG. 13 .
The material property prediction presentation screen 332 consists of display elements of a material property prediction value display frame 1301, a parameter importance display frame 1310, a material information detail display frame 1320, and a completion button 710, as shown in FIG.

In the material property predicted value display frame 1301, the values of the material ID 601, the material property 602, the material property predicted value 608 of the material data visualization table 250, and the material name 502 of the material data table 240 are displayed. A material property display field 1302 displayed in a material property predicted value display frame 1301 is displayed using the value of the material property 602 or, if there is no such value, the value of the material property predicted value 608 . At this time, when the material property predicted value 608 is displayed, it can be displayed, for example, underlined so that it can be distinguished from it. Also, when the user selects one of the displayed materials, the details of that material can be displayed in the material information detail display frame 1320 .

The parameter importance display frame 1310 displays the parameter importance corresponding to the material of the material ID with reference to the material parameter 603 of the material data visualization table 250 and the parameter importance 1102 of the parameter importance table 260 .

As described above, the material property prediction device of the present embodiment makes it possible to easily incorporate the user's knowledge into the prediction when predicting material properties for screening of experimental plans. This makes it easier to plan experiments, which in turn makes it possible to develop good materials with fewer experiments.

[Embodiment 2]
A second embodiment according to the present invention will be described below with reference to FIGS. 14 to 17. FIG.
The first embodiment relates to a material property prediction device that enables material prediction that makes use of the user's knowledge by editing the material data points displayed on the material data 2D display screen 312 . In the first embodiment, the material data is visualized in a two-dimensional space, but since the sense of distance cannot be grasped, it may be difficult to edit. In this embodiment, by associating some existing image with two-dimensional coordinates, the image can be displayed so that the sense of distance can be grasped.

Therefore, the material property prediction device of this embodiment is a material property prediction device having the same idea as that of the first embodiment, and by displaying a background drawing on the screen of the material data 2D display screen 312, the relationship between data can be grasped. It's made to make it easier.

In the following description of the material property prediction device of the present embodiment, differences from the first embodiment will be mainly described.
First, the configuration of the material property prediction device of Embodiment 2 will be described with reference to FIG.
The material property prediction device 101 of the present embodiment has a material visualization background table 1401 in addition to the material property prediction device 101 shown in FIG. 1 of the first embodiment. Then, the 2D visualization unit 115 constructs a screen using this material visualization background table 1401 .

Next, the data structure of the material visualization background table 1401 will be described using FIG.
The material visualization background table 1401 is a table that stores the image data of the background image of the material data display frame 701 of the material data 2D display screen 312. As shown in FIG. , upper right X coordinate 1603 , upper right Y coordinate 1604 , and background image data 1605 .

A lower left X coordinate 1601, a lower left Y coordinate 1602, an upper right X coordinate 1603, and an upper right Y coordinate 1604 represent coordinate points of the corresponding background image data 1605, respectively.

Next, the material data visualization processing of the second embodiment will be described with reference to FIG. 16 .
The difference from the first embodiment is that after the X and Y coordinates of the material data point are determined by the dimension compression of the material parameter by the dimension compression unit 113 without teacher data, the data of the material visualization background table 1401 is referred to and the coordinate origin and The point is that there is a process (S110) for matching the scales.

The 2D visualization unit 115 obtains the center of the coordinates of the background image whose image data is stored in the material visualization background table 1401, obtains the center of gravity of the X coordinate 606 and the Y coordinate 607 of the material data visualization table 250, and obtains the difference between the two. to calculate By subtracting this difference from the X coordinate 606 and Y coordinate 607 of the material data visualization table 250, the coordinates of the background image and the material data visualization table 250 can be matched. As for this correction method, any method may be used as long as it is a method that appropriately overlaps the background image.

Next, the material data 2D display screen of the second embodiment will be described with reference to FIG.
A map of Japan is superimposed and displayed as a background image in the material data display frame 701 of the material data 2D display screen of the present embodiment. In this way, by displaying a background image that is familiar and intuitive to the user, it is possible to intuitively grasp the distance between the material data.

Reference numerals 101: material property prediction device 112: prediction evaluation unit 113: no teacher data dimension compression unit 115: 2D visualization unit 116: semi-supervised data presence dimension compression unit 117: material property prediction presentation unit 240: material data table 250: material data visualization Table 260: Parameter importance table 270: Material/coordinate mapping table

Claims (12)

A material property prediction device for predicting and displaying material properties,
a material data table storing material properties and material parameters associated with the material;
a material data visualization table storing dimensionally compressed information of the material parameters as two-dimensional coordinates;
holding a parameter importance table that stores parameter importance associated with each of the material parameters;
A dimension compression unit without teacher data that stores the material parameters as two-dimensional coordinates by dimension compression in the material data visualization table ;
a 2D visualization unit that two-dimensionally displays the two-dimensional coordinates stored in the material data visualization table as material data points and receives position changes of the material data points;
Based on the received position change information of the material data point, the changed material data point is set as an objective variable, and the material parameter of the material data visualization table corresponding to the changed material data point and the parameter importance for the material parameter are determined. Calculate the dimension compression function and the parameter importance of the dimension compression function with respect to the material parameter by regression analysis using explanatory variables, update the parameter importance stored in the parameter importance table with the calculated parameter importance, and position Input material parameters of changed material data points and unchanged material data points , output as two-dimensional coordinates dimensionally compressed by the dimension compression function, and store the output two-dimensional coordinates in the material data visualization table. a dimensional compression unit with semi-supervised data that updates as stored two-dimensional coordinates;
and a material property prediction presenting unit for predicting material properties based on the updated parameter importance . 2. A material property prediction apparatus according to claim 1, further comprising a prediction evaluation unit for evaluating material properties by a material property prediction function that returns a predicted value of material properties from said material parameters. 3. A material property prediction apparatus according to claim 2 , wherein said material property prediction function uses a Euclidean distance multiplied by a weighting factor for each dimension. The prediction evaluation unit calculates the prediction accuracy of material properties,
3. The material property prediction device according to claim 2 , wherein the 2D visualization unit displays information about the prediction accuracy. 2. A material property prediction apparatus according to claim 1, wherein said material property prediction presentation unit displays information on each material property. further maintaining a background image for evaluating distances between said material data points ;
2. The material property prediction device according to claim 1, wherein the 2D visualization unit displays the background image on the background of the material data points. A material property prediction method for predicting and displaying material properties by a material property prediction device ,
The material property prediction device is
a material data table storing material properties and material parameters associated with the material;
a material data visualization table storing dimensionally compressed information of the material parameters as two-dimensional coordinates;
holding a parameter importance table that stores parameter importance associated with each of the material parameters;
A dimensional compression step without teacher data in which the material property prediction device stores the material parameters as two-dimensional coordinates by dimensional compression in the material data table;
a 2D visualization step of two-dimensionally displaying the two-dimensional coordinates stored in the material data visualization table as material data points and accepting position changes of the material data points;
Based on the received position change information of the material data point, the changed material data point is set as an objective variable, and the material parameter of the material data visualization table corresponding to the changed material data point and the parameter importance for the material parameter are determined. Calculate the dimension compression function and the parameter importance of the dimension compression function with respect to the material parameter by regression analysis using explanatory variables, update the parameter importance stored in the parameter importance table with the calculated parameter importance, and position The material parameters of the material data points whose positions have been changed and the material data points whose positions have not been changed are output as two-dimensional coordinates dimensionally compressed by the input dimension compression function , and the output two-dimensional coordinates are stored in the material data visualization table. a dimensionality reduction step with semi-supervised data that updates as two-dimensional coordinates of
and a material property prediction presenting step of predicting and presenting material properties based on the updated parameters . 8. The material property prediction method according to claim 7, further comprising a prediction evaluation step of evaluating the material property using a material property prediction function that returns a predicted value of the material property from the material parameter. 9. The material property prediction method according to claim 8 , wherein said material property prediction function uses a Euclidean distance multiplied by a weighting factor for each dimension. In the prediction evaluation step, calculating the prediction accuracy of the material properties,
9. The method of predicting material properties according to claim 8 , wherein information about the prediction accuracy is displayed in the 2D visualization step. Furthermore, having a material property prediction presentation step,
8. The material property prediction method according to claim 7 , wherein in said material property prediction presentation step, information on each material property is displayed. further maintaining a background image for evaluating distances between said material data points;
8. The method of claim 7, wherein the 2D visualization step displays the background image behind the material data points.

Images