This application is based upon and claims the benefit of priority from the prior Japanese Patent Application No. 2022-147965, filed on Sep. 16, 2022, the entire contents of which are incorporated herein by reference.
Embodiments described herein relate generally to an information processing apparatus and an information processing method.
Sparse modeling for extracting an explanatory variable having a large degree of influence on an objective variable among a large number of explanatory variables has been known.
There may be another explanatory variable having a large degree of influence on at least a part of the explanatory variables among the plurality of extracted explanatory variables, and there is a concern that the explanatory variable having a large degree of influence on the objective variable cannot be extracted without omission in existing sparse modeling.
In addition, in a case where the explanatory variables are similar to each other, there is a concern that the explanatory variable having a large degree of influence on the objective variable cannot be appropriately extracted in the existing sparse modeling.
In general, according to the embodiment, an information processing apparatus comprising processing circuitry. The processing circuitry is configured to acquire objective variables and explanatory variables which are regression analysis targets, extract a plurality of first explanatory variables having a high degree of influence on the objective variable from among the explanatory variables by sparse modeling using a first regression equation, and extract a second explanatory variable having a high degree of influence on the plurality of first explanatory variables by sparse modeling using a second regression equation. Hereinafter, information processing apparatuses of the present disclosure will be described with reference to the drawings.
The database unit 2 stores objective variables and explanatory variables which are regression analysis targets. The database unit 2 also serves as an acquisition unit, and the objective variable and the explanatory variable acquired from the database unit 2 are supplied to the first processing unit 3. In addition, the explanatory variable acquired from the database unit 2 is also supplied to the second processing unit 4.
Instead of the database unit 2, the regression analysis target may be monitored and a monitor image or the like may be supplied as the objective variable or the explanatory variable to the first processing unit 3 or the second processing unit 4. For example, in a case where defect analysis in a manufacturing process of a semiconductor wafer is performed, the semiconductor wafer may be captured in the manufacturing process, and the captured image may be supplied as the explanatory variable to the first processing unit 3 or the second processing unit 4, for example.
The first processing unit 3 extracts a plurality of first explanatory variables having a high degree of influence on the objective variable by sparse modeling using a first regression equation. The number of first explanatory variables extracted by the first processing unit 3 may be two or more. The plurality of first explanatory variables extracted by the first processing unit 3 are sent to the second processing unit 4.
The second processing unit 4 extracts a second explanatory variable having a high degree of influence on the plurality of first explanatory variables by sparse modeling using a second regression equation. The number of second explanatory variables extracted by the second processing unit 4 may be one or more, and there may be a plurality of second explanatory variables.
A specific method of the sparse modeling adopted by the first processing unit 3 and the second processing unit 4 is not limited to a particular method. Hereinafter, an example in which the first processing unit 3 and the second processing unit 4 use sparse modeling by multi-task Lasso will be described.
The first processing unit 3 extracts a plurality of first explanatory variables by using a first regression coefficient indicating a degree of influence on the objective variable. The second processing unit 4 calculates the second explanatory variable by using a second regression coefficient indicating a degree of influence on the plurality of first explanatory variables.
As a more specific example, the first processing unit 3 includes a first degree-of-influence update unit (first update unit) 11, a first regression error calculation unit (first calculation unit) 12, a first degree-of-influence density calculation unit (second calculation unit) 13, a first score calculation unit (third calculation unit) 14, and a first convergence determination unit (first determination unit) 15.
The first degree-of-influence update unit 11 updates the first regression coefficient indicating the degree of influence of the first explanatory variable on the objective variable based on the objective variable and the first explanatory variable.
The first regression error calculation unit 12 calculates a regression error indicating accuracy in a case where the objective variable is regressed from the first regression coefficient updated by the first degree-of-influence update unit 11 and the first explanatory variable.
The first degree-of-influence density calculation unit 13 calculates the number (density) of first regression coefficients such that the regression error calculated by the first regression error calculation unit 12 becomes small.
The first score calculation unit 14 calculates a score that is a value of the first regression equation based on a calculation result of the first degree-of-influence density calculation unit 13 and a calculation result of the first regression error calculation unit 12.
The first convergence determination unit 15 determines whether or not the score calculated by the first score calculation unit 14 satisfies a convergence condition. Since it is desirable that the score is as small as possible, the first convergence determination unit 15 may determine that the convergence condition is satisfied when the score is less than a predetermined threshold.
The first processing unit 3 repeats kinds of processing of the first degree-of-influence update unit 11, the first regression error calculation unit 12, the first degree-of-influence density calculation unit 13, and the first score calculation unit 14 until the first convergence determination unit 15 determines that the score satisfies the convergence condition.
The second processing unit 4 includes a second degree-of-influence update unit (second update unit) 21, a second regression error calculation unit (fourth calculation unit) 22, a second degree-of-influence density calculation unit 23 (fifth calculation unit), a second score calculation unit (sixth calculation unit) 24, and a second convergence determination unit 25 (second determination unit).
The second degree-of-influence update unit 21 updates the second regression coefficient indicating the degree of influence of the second explanatory variable on each of the plurality of first explanatory variables.
The second regression error calculation unit 22 calculates a regression error indicating accuracy in a case where the corresponding first explanatory variable is regressed from the second regression coefficient updated by the second degree-of-influence update unit 21 and the second explanatory variable.
The second degree-of-influence density calculation unit 23 calculates the number (density) of second regression coefficients such that the regression error calculated by the second regression error calculation unit 22 becomes small.
The second score calculation unit 24 calculates a score that is a value of the second regression equation based on a calculation result of the second degree-of-influence density calculation unit 23.
The second convergence determination unit 25 determines whether or not the score calculated by the second score calculation unit 24 satisfies a convergence condition. In the present embodiment, since it is desirable that the score is as small as possible, the second convergence determination unit 25 may determine that the convergence condition is satisfied when the score is less than a predetermined threshold.
The second processing unit 4 repeats kinds of processing of the second degree-of-influence update unit 21, the second regression error calculation unit 22, the second degree-of-influence density calculation unit 23, and the second score calculation unit 24 until the second convergence determination unit 25 determines that the score satisfies the convergence condition.
In the information processing apparatus 1 of
The first degree-of-influence reception unit 16 receives the plurality of first regression coefficients (degrees of influence) corresponding to the plurality of first explanatory variables extracted by the first processing unit 3. The second degree-of-influence reception unit 26 receives the second regression coefficient (degree of influence) corresponding to the second explanatory variable extracted by the second processing unit 4. The defect factor selection unit 5 sequentially selects the plurality of first regression coefficients corresponding to the plurality of first explanatory variables extracted by the first processing unit 3, and supplies the plurality of first regression coefficients to the second degree-of-influence update unit 21 in the second processing unit 4.
The defect factor identification unit 6 identifies, as defect factors, the plurality of first explanatory variables extracted by the first processing unit 3 and the second explanatory variables extracted by the second processing unit 4. The defect factor display unit 7 displays the defect factor identified by the defect factor identification unit 6 on a display unit of a display device (not illustrated) or various electronic devices. The electronic device is a PC, a tablet, a smartphone, or the like. The degree-of-influence display unit 8 displays a plurality of degrees of influence corresponding to the plurality of first explanatory variables and degrees of influence corresponding to the second explanatory variables on the display unit of the display device or the electronic device. The defect factor display unit 7 and the degree-of-influence display unit 8 may display the defect factor and the degrees of influence on the display unit of the identical display device or electronic device.
For example, the first processing unit 3 calculates a first regression coefficient β that minimizes a score of the first regression equation represented in the following Equation (1), and calculates components of a plurality of first explanatory variables X corresponding to components of the large first regression coefficient β. Specifically, the components (indices) I=(i1, i2, . . . ) of the plurality of first explanatory variables X are obtained.
A first term on a right side of Equation (1) is a norm of a difference between a value obtained by multiplying the first explanatory variable X by the first regression coefficient β and the objective variable Y, and more specifically, a value obtained by averaging the sum of squares of the above-described difference is calculated. A second term on the right side is a regularization term obtained by multiplying a norm of the first regression coefficient β by a first regularization coefficient λ.
For i∈I, the second processing unit 4 sets components other than an update target component (hereinafter, component i) among the plurality of first explanatory variables X as X′, obtains a second regression coefficient β′ that minimizes a second regression equation L′(β) represented in the following Equation (2), and obtains components (indices) Ii=(ji, 1, ji, 2, . . . ) of the second explanatory variable X′ corresponding to the component of the large second regression coefficient β′.
A first term on a right side of Equation (2) is a norm of a difference between a value obtained by multiplying the second explanatory variable X′ by the second regression coefficient β′ and a first explanatory variable Xi, and more specifically, is a value obtained by averaging the sum of squares of the above-described difference. A second term on the right side is a normalization term obtained by multiplying a norm of the second regression coefficient β′ by a second regularization coefficient λ′.
The information processing apparatus 1 of
First, the first processing unit 3 reads out the objective variable and the explanatory variable registered in the database unit 2 (step S1). In step S1, a larger number of explanatory variables than the plurality of first explanatory variables finally extracted by the first processing unit 3 are read out.
The first degree-of-influence update unit 11 updates the first regression coefficient indicating the degree of influence of the explanatory variable on the objective variable based on the objective variable and the explanatory variable read out in step S1 (step S2).
Subsequently, the first regression error calculation unit 12 calculates the regression error indicating the accuracy in a case where the objective variable is regressed from the first regression coefficient updated in step S2 and the explanatory variable (step S3). Subsequently, the first degree-of-influence density calculation unit 13 calculates the number (density) of first regression coefficients such that the regression error calculated in step S3 becomes small (step S4). By the processing of step S4, the number of the plurality of extracted first explanatory variables and the number of corresponding first regression coefficients are adjusted.
Subsequently, the first score calculation unit 14 calculates the score of the first regression equation based on the calculation result of step S4 (step S5). Subsequently, the first convergence determination unit 15 determines whether or not the score satisfies a predetermined convergence condition (step S6). Since a smaller score is generally more desirable, the convergence condition is set to, for example, an upper limit of an acceptable score.
When it is determined in step S6 that the score does not satisfy the convergence condition, the kinds of processing of steps S2 to S6 are repeated. When it is determined in step S6 that the score satisfies the convergence condition, a plurality of first regression coefficients corresponding to the extracted plurality of first explanatory variables are sent to the first degree-of-influence reception unit 16 (step S7). The plurality of extracted first explanatory variables correspond to a plurality of defect factors.
The defect factor selection unit 5 sequentially selects one of the plurality of first regression coefficients corresponding to the plurality of defect factors (step S8). That is, in step S8, selection is performed for each defect factor. Subsequently, the corresponding explanatory variable (second explanatory variable) is acquired from the database unit 2 (step S9).
The kinds of subsequent processing in steps S10 to S15 are performed for each first regression coefficient corresponding to the defect factor selected in step S8. First, the second degree-of-influence update unit 21 updates a first coefficient and a second coefficient indicating the degree of influence of the second explanatory variable on the selected first explanatory variable (step S10). The first coefficient is a regression coefficient of the second regression equation, and the second coefficient is a coefficient indicating a multicollinearity relationship. The first coefficient corresponds to the second regression coefficient β′ of Equation (2), and the second coefficient corresponds to the second regularization coefficient λ′ of Equation (2). In the present specification, the first coefficient and the second coefficient correspond to the second regression coefficient.
Subsequently, the second regression error calculation unit 22 calculates a regression error indicating the accuracy in a case where the selected first explanatory variable is regressed from the second regression coefficient updated in step S10 and the second explanatory variable (step S11). Subsequently, the second degree-of-influence density calculation unit 23 calculates the number (density) of second coefficients such that the regression error calculated in step S11 becomes small (step S12).
Subsequently, the second score calculation unit 24 calculates the score of the second regression equation based on the calculation result of step S12 (step S13). Subsequently, the second convergence determination unit 25 determines whether or not the score satisfies a predetermined convergence condition (step S14).
When it is determined in step S14 that the score does not satisfy the convergence condition, the kinds of processing of steps S10 to S14 are repeated. When it is determined in step S14 that the score satisfies the convergence condition, the extracted second explanatory variable is transmitted to the second degree-of-influence reception unit 26 (step S15).
When the kinds of processing of steps S10 to S15 are completed for the plurality of defect factors, it is determined whether or not the kinds of processing of steps S10 to S15 are performed for all the defect factors (step S16). When there remains a defect factor that has not yet been processed, the kinds of processing of steps S8 to S16 are repeated.
When the kinds of processing of steps S10 to S15 are completed for all the defect factors, the defect factor identification unit 6 identifies, as defect factors, the plurality of first explanatory variables obtained in the processing of the first processing unit 3 (steps S2 to S7) and the second explanatory variable obtained in the processing of the second processing unit 4 (steps S10 to S15) (step S17).
Subsequently, the defect factor display unit 7 displays the defect factor identified in step S15 (step S18). In addition, the degree-of-influence display unit 8 displays the degrees of influence respectively corresponding to the plurality of first explanatory variables and the second explanatory variables corresponding to the defect factors (step S19).
Consequently, according to the first embodiment, not only the plurality of first explanatory variables having the high degree of influence on the objective variable Y can be extracted, but also the second explanatory variable having the high degree of influence on each of the first explanatory variables can be extracted, and all the explanatory variables having the high degree of influence on the objective variable Y can be extracted without omission.
Although the second processing unit 4 alone does not have a high degree of influence on the objective variable Y, the second processing unit can also extract a second explanatory variable having a high degree of influence on the first explanatory variable by combining the plurality of the second explanatory variables.
Specific examples in which the degree of influence on the first explanatory variable is increased by combining the plurality of second explanatory variables include, in addition to an example in which the plurality of second explanatory variables are added (order 1), an example in which the plurality of second explanatory variables are subtracted (order 2), an example in which a product-sum calculation of the plurality of second explanatory variables is performed (order 4), and an example in which a product-difference calculation of the plurality of second explanatory variables is performed (order 5).
The information processing apparatus 100 of
Since each block other than the semiconductor observation unit 39 and the similar feature value calculation unit 40 performs processing similar to the block having the identical name of
The similar feature value calculation unit 40 calculates a similar feature value similar to a feature value of the explanatory variable, and extracts an explanatory variable similar to the explanatory variable.
Since the information processing apparatus 100 according to the comparative example illustrated in
As described above, in the first embodiment, after the sparse modeling for extracting the plurality of first explanatory variables having a high degree of influence on the objective variable is performed, since the sparse modeling for extracting the second explanatory variable having a high degree of influence on the plurality of first explanatory variables is performed, the explanatory variables having a high degree of influence on the objective variable can be extracted substantially without omission.
In the first embodiment, since sparse modeling using two types of regression equations is performed, there is a concern that it takes time to obtain a regression analysis result. In a second embodiment to be described below, a regression analysis result is obtained by one regression analysis.
A regression equation used by an information processing apparatus 1a according to the second embodiment includes a term related to an explanatory variable having a multicollinearity relationship, as represented in Equation (3). λ, λ2, and λ3 in Equation (3) are regularization coefficients.
A first term on a right side of Equation (3) is a norm of a difference between a value obtained by multiplying the first explanatory variable X by the first coefficient β and the objective variable Y, and more specifically, corresponds to a value obtained by averaging the sum of squares of the above-described difference. A second term on the right side is a value obtained by multiplying a norm of the first coefficient β by the first regularization coefficient λ, and more specifically, is a sum of absolute values of values obtained by multiplying the norm of the first coefficient β by the first regularization coefficient λ. A third term on the right side is an absolute value of a value obtained by multiplying a norm of a value obtained by multiplying the second explanatory variable X by a second coefficient C by a second regularization coefficient λ2. A fourth term on the right side is an absolute value of a value obtained by multiplying a norm of the second coefficient C by a third regularization coefficient λ3.
The multicollinearity means a case where there is a plurality of collinearity having a linear relationship (primary dependency) between the explanatory variables. When there are a plurality of explanatory variables having a multicollinearity relationship, there is a concern that the explanatory variable having a high degree of influence on the objective variable cannot be appropriately extracted. Therefore, in Equation (3), regularization terms (the third term and the fourth term on the right side) related to multicollinearity are provided, and sparse modeling is performed in consideration of explanatory variables having a multicollinearity relationship.
In the present specification, β of Equation (3) is referred to as the first coefficient, and C is referred to as the second coefficient. The first coefficient β is the regression coefficient. The second coefficient C is a coefficient indicating a multicollinearity relationship, and is a coefficient indicated by, for example, a p-dimensional×p-dimensional square matrix as illustrated in
As illustrated in
The information processing apparatus 1a according to the second embodiment obtains the second coefficient C such that the third term on the right side of Equation (3) becomes zero and the fourth term on the right side becomes as small as possible.
The processing unit 41 of
The degree-of-influence update unit 42 updates a regression coefficient indicating the degree of influence of the first explanatory variable on the objective variable based on the first explanatory variable extracted from the explanatory variable and the objective variable.
The regression error calculation unit 43 calculates a regression error indicating a regression error in a case where the objective variable is regressed from the regression coefficient updated by the degree-of-influence update unit 42 and the first explanatory variable.
The degree-of-influence density calculation unit 44 calculates the number (density) of regression coefficients such that the regression error calculated by the regression error calculation unit 43 becomes small.
The multicollinearity error calculation unit 45 calculates an error of multicollinearity from the second explanatory variable and the second coefficient. More specifically, the multicollinearity error calculation unit 45 obtains the second explanatory variable and the second coefficient C in the third term on the right side of Equation (3) such that the score of the regression equation is minimized.
The multicollinearity density calculation unit 46 calculates the number (density) of second coefficients. More specifically, the multicollinearity density calculation unit 46 obtains the second coefficient C of the fourth term on the right side of Equation (3) such that the score of the regression equation is minimized.
As described above, since the second coefficient C is a value related to the first coefficient β, not only the second coefficient C but also the first coefficient β is obtained by performing the kinds of processing of the multicollinearity error calculation unit 45 and the multicollinearity density calculation unit 46.
The score calculation unit 47 calculates the score of the regression equation represented in Equation (3) based on the calculation result of the multicollinearity density calculation unit 46.
The convergence determination unit 48 determines whether or not the score obtained by the score calculation unit 47 satisfies a convergence condition. Since it is desirable that the score is as small as possible, the convergence determination unit 48 may determine that the convergence condition is satisfied when the score is less than a predetermined threshold.
The processing unit 41 of
In addition, the information processing apparatus 1a of
The defect factor identification unit 6 identifies, as defect factors, the first explanatory variable and the second explanatory variable corresponding to the first coefficient β and the second coefficient C adjusted by the degree-of-influence adjustment unit 51. Processing operations of the defect factor display unit 7 and the degree-of-influence display unit 8 are similar to the processing operations of
Subsequently, the degree-of-influence update unit 42 updates the regression coefficient indicating the degree of influence of the first explanatory variable on the objective variable based on the first explanatory variable extracted from the explanatory variable and the objective variable (step S22).
Subsequently, the regression error calculation unit 43 calculates a regression error indicating the accuracy in a case where the objective variable is regressed from the regression coefficient updated by the degree-of-influence update unit 42 and the first explanatory variable (step S23).
Subsequently, the degree-of-influence density calculation unit 44 calculates the number (density) of regression coefficients such that the regression error calculated by the regression error calculation unit 43 becomes small (step S24).
Subsequently, the multicollinearity error calculation unit 45 calculates an error of multicollinearity from the second explanatory variable and the second coefficient (step S25). More specifically, the multicollinearity error calculation unit 45 obtains the second explanatory variable and the second coefficient C in the third term on the right side of Equation (3) such that the score of the regression equation is minimized.
Subsequently, the multicollinearity density calculation unit 46 calculates the number (density) of second coefficients (step S26). More specifically, the multicollinearity density calculation unit 46 obtains the second coefficient C of the fourth term on the right side of Equation (3) such that the score of the regression equation is minimized.
Subsequently, the score calculation unit 47 calculates the score of the regression equation represented in Equation (3) based on the calculation result of the multicollinearity density calculation unit 46 (step S27).
Subsequently, the convergence determination unit 48 determines whether or not the score obtained by the score calculation unit 47 satisfies a convergence condition (step S28). When it is determined that the score does not satisfy the convergence condition, the kinds of processing of steps S22 to S28 are repeated. When it is determined that the score satisfies the convergence condition, the degree-of-influence adjustment unit 51 performs processing of adjusting the number of first coefficients 13 and second coefficients C to an appropriate number in a case where the number of first coefficients 13 and second coefficients C extracted by performing the kinds of processing of the multicollinearity error calculation unit 45 and the multicollinearity density calculation unit 46 described above are too large (step S29).
Subsequently, the defect factor identification unit 6 identifies, as defect factors, the first explanatory variable and the second explanatory variable corresponding to the first coefficient β and the second coefficient C adjusted by the degree-of-influence adjustment unit 51 (step S30).
Subsequently, the defect factor display unit 7 displays the defect factor (step S31). In addition, the degree-of-influence display unit 8 displays the degree of influence including the first coefficient β and the second coefficient C (step S32).
As described above, in the second embodiment, the plurality of first explanatory variables having a high degree of influence on the objective variable and the second explanatory variables having a high degree of influence on the plurality of first explanatory variables can be extracted by sparse modeling in one stage. Thus, as compared with the first embodiment in which the sparse modeling is performed in two stages, the processing operation of the information processing apparatus 1a can be simplified, and the explanatory variable can be extracted more quickly.
In addition, in the second embodiment, since the regression equation used for sparse modeling includes a penalty term (regularization term) for the explanatory variable having the multicollinearity relationship, it is possible to quickly extract a desired explanatory variable without being influenced by the explanatory variable having the multicollinearity relationship.
The use of the information processing apparatuses 1 and 1a according to the first and second embodiments described above is not limited, but can be used, for example, for defect analysis in the manufacturing process of the semiconductor device. The semiconductor device is manufactured by performing a large number of processes such as film formation, exposure, and etching on a semiconductor wafer. There is a concern that a defect occurs during the processing of any process, and it is possible to identify a factor of the defect and take measures against the defect by analyzing what kind of defect has occurred in which process. Thus, yield is improved.
The semiconductor wafer can be classified into four types of reference images as illustrated in
In a case where the defect analysis of the semiconductor wafer is performed, it is conceivable to detect, by image analysis, which of the reference images IM1 to IM4 corresponds to an image obtained by visualizing, as the objective variable, an electrical characteristic value, a measured value, or the like of the semiconductor wafer to be subjected to defect analysis. However, the identical semiconductor wafer may include defects of a plurality of different forms. In a case where a combination of a plurality of different defects is taken into consideration when the number of defect forms increases, the number of reference images becomes very large, and it takes a lot of time to classify the image to be subjected to defect analysis. In addition, in a case where a defect occurrence rate is greatly different for each defect form, there is a concern that a true cause of the defect is not accurately determined for a defect form having a low defect occurrence rate.
In contrast, since the plurality of first explanatory variables having a high degree of influence on the objective variable and the second explanatory variables having a high degree of influence on the plurality of first explanatory variables can be quickly and accurately extracted by performing the sparse modeling by Equation (1) and Equation (2) or the sparse modeling by Equation (3) described above, the true cause of the defect can be accurately determined.
According to the first embodiment or the second embodiment, as illustrated in
In
In addition, as illustrated in
In contrast, in the information processing apparatus 1 or 1a according to the first or second embodiment, as illustrated in
As described above, the information processing apparatus 1 according to the third embodiment can accurately extract the plurality of first explanatory variables having a high degree of influence on the objective variable and the second explanatory variables having a high degree of influence on the plurality of first explanatory variables in various fields having a large number of explanatory variables that can influence the objective variable.
At least a part of the information processing apparatuses 1 and 1a described in the above-described embodiments may be achieved by hardware or software. In a case where the at least a part thereof is achieved by software, a program that achieves at least a part of the functions of the information processing apparatuses 1 and 1a may be stored in a recording medium such as a flexible disk or a CD-ROM, and may be read and executed by a computer. The recording medium is not limited to an attachable and detachable medium such as a magnetic disk or an optical disk, and may be a fixed recording medium such as a hard disk device or a memory.
In addition, the program that achieves at least a part of the functions of the information processing apparatuses 1 and 1a may be distributed via a communication line (including wireless communication) such as the Internet. The program may be distributed via a wired line or a wireless line such as the Internet in a state of being encrypted, modulated, and compressed or in a state of being stored in the recording medium.
The above-described examples may be configured as follows.
(1) An information processing apparatus comprising processing circuitry, the processing circuitry configured to:
While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel devices and methods described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the embodiments described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modification as would fall within the scope and spirit of the inventions.
Number | Date | Country | Kind |
---|---|---|---|
2022-147965 | Sep 2022 | JP | national |