Operation program, operation method and information processing apparatus
By using spectral data of elemental substances as explanatory variables and applying regularization in multivariate regression analysis, the complexity of spectral data analysis is reduced, achieving accurate and efficient composition ratio determination.
Patent Information
- Application Number
- JP2023071539
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-25
- Publication Date
- 2024-11-07
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The process of analyzing spectral data becomes complicated as the number of elemental substances increases, requiring numerous samples and spectra, especially in multivariate regression analysis, which necessitates a large number of spectral data and complex preparation.
Utilizing spectral data of each elemental substance as explanatory variables and applying regularization in multivariate regression analysis to simplify the process, preventing overfitting and improving accuracy by selectively reducing regression coefficients.
Simplifies the analysis process while maintaining high accuracy by reducing the number of necessary samples and spectra, and enhancing the precision of composition ratio determination.
Smart Images

Figure 2024157277000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a computing program, a computing method, and an information processing device. [Background technology]
[0002] Techniques have been disclosed for quantifying component substances from spectral data obtained by analyzing a substance to be analyzed (see, for example, Patent Documents 1 to 4). [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2019-215356 A [Patent Document 2] JP 2020-201174 A [Patent Document 3] Patent Publication No. 2021-9135 [Patent Document 4] Special Publication No. 2010-520471 Summary of the Invention [Problem to be solved by the invention]
[0004] For example, in order to analyze spectral data, a plurality of samples with known composition ratios are prepared, spectral data for the prepared samples are acquired, and a calibration curve is created. However, in this method, it is necessary to prepare a plurality of samples with different composition ratios to acquire spectral data, and as the number of element substances increases, the number of samples and spectra to be prepared also increases, making the process more complicated.
[0005] In one aspect, the present invention aims to provide a calculation program, a calculation method, and an information processing device that can simplify a process. [Means for solving the problem]
[0006] In one aspect, the calculation program causes a computer to execute a process of performing regularization using spectral data of each of a plurality of component substances contained in a target substance as explanatory variables and spectral data obtained as an analysis result for the target substance as a response variable, and identifying a composition ratio of the plurality of component substances in the target substance. Effect of the Invention
[0007] The process can be simplified. [Brief description of the drawings]
[0008] [Figure 1] This is an example with two-dimensional explanatory variables. [Diagram 2] FIG. 13 is a diagram illustrating contours of the regularization term in L2 regularization. [Diagram 3] FIG. 13 is a diagram illustrating contours of the regularization term in L1 regularization. [Figure 4] (a) to (d) are IR spectrum data obtained by simulation. [Diagram 5] The regression analysis is shown in the figure. [Figure 6] (a) to (e) show the results of regression analysis using regularization. [Figure 7] FIG. 2A is a functional block diagram illustrating an example of an overall configuration of an information processing device, and FIG. 2B is a block diagram illustrating an example of a hardware configuration of the information processing device. [Figure 8] 11 is a flowchart illustrating an example of an operation of the information processing device. [Figure 9] 11 is a flowchart illustrating an example of an operation of the information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] Compositional and quantitative analysis is the analysis of the composition (types of constituent elements) and their abundance ratio (composition ratio) in a substance, and for example, uses the peak position of the spectrum data and its intensity and area. Several samples with different amounts of the target element substance are prepared in advance, and when the amount of the element substance changes, the spectral position that changes is found and identified. Then, a calibration curve formula is obtained from the relationship between the amount of the target element substance and the spectral intensity or area at the identified position. Next, the amount of the target element substance is obtained by applying the spectral intensity or area of the sample, which is the target substance for compositional and quantitative analysis, to the calibration curve formula. This method is a simple regression analysis that creates a regression line using a single variable. However, if the spectral data used is affected by other element substances, a good calibration curve cannot be obtained, and it is necessary to separate the target element substance from the interfering components in advance.
[0010] For this reason, in recent years, the use of multivariate regression analysis, which creates a regression line using multiple variables, has been proposed. There are various methods for multivariate regression analysis, such as multiple regression analysis, principal component analysis, factor analysis, and cluster analysis.
[0011] Multivariate regression analysis has three advantages over simple regression analysis: (I) there is no need to determine the position of the peaks, (II) quantification is possible even if the peaks overlap in a complex manner, and (III) multiple elemental substances can be quantified at once. However, for (I), more accurate results can be obtained by narrowing down the analysis range to some extent according to the information on the elemental substance of interest or by limiting it to an arbitrary fractional range that does not include noise components.
[0012] In multivariate regression analysis, the method using PLS (Partial Least Squares) regression that utilizes principal component analysis and multiple regression analysis (PLS method) is recommended in terms of accuracy as mentioned above. The PLS method is carried out in two steps. In the first step, multivariate data is subjected to principal component analysis to summarize all variables into "principal components" (approximately 1 to 20 variables), and in the second step, multiple principal components are combined using multiple regression analysis to derive a single optimal variable. In multiple regression analysis, a population is created by preparing a large amount of spectral data of samples in which the concentration of the target element substance is known, and a weighting coefficient b is calculated so that the residual (the difference between the known concentration value and the concentration value calculated from the spectral data using the following formula (1)) is minimized when the spectral data in the population is applied to the following formula (1): i In this two-step analysis, the most important thing is how to combine the "principal components" in the intermediate step.
number
[0013] As described above, multivariate regression analysis has the three advantages (I) to (III) over simple regression analysis, but in order to obtain high accuracy, it is necessary to narrow down the variables (target peaks) using principal component analysis. As with simple regression analysis, the objective variable is concentration (amount), the explanatory variable is spectral data, and the goal is to obtain a highly accurate calibration curve by using spectral data corresponding to multiple concentrations (amounts). To achieve this, a large number of spectral data corresponding to known concentrations must be prepared; for example, 3 to 10 pieces are required for simple regression analysis, and 20 or more pieces are required for multivariate analysis. Although the use of multivariate analysis has several advantages over simple regression analysis, the process is complicated, and simplification of the process is an issue.
[0014] In the following embodiment, an example in which a complicated process can be simplified will be described. EXAMPLES
[0015] First, the principle of the first embodiment will be described.
[0016] In this embodiment, the target substance includes a plurality of component substances. The plurality of component substances includes at least a compound of two or more kinds of components or a mixture of two or more kinds of components. Therefore, for example, the target substance is a compound obtained by reacting two or more kinds of components, a mixture in which two or more kinds of components are simply mixed, etc.
[0017] First, the spectral data of each of the multiple elemental substances contained in the substance to be analyzed is prepared. The spectral data of each of the multiple elemental substances can be obtained in advance by analyzing each of the multiple elemental substances. Next, the composition ratio of the multiple elemental substances is obtained by multiple regression analysis using regularization. For regularization, combinatorial optimization techniques such as digital annealing can be used. Here, regularization is a method used in machine learning to prevent overfitting. In machine learning, a computer learns from training data prepared in advance and creates a model. Generally, training is performed to minimize the loss function for the training data, but this alone generates a complex model that is finely fitted, and overfitting occurs. Overfitting is a phenomenon in machine learning where the model fits the training data too faithfully, resulting in a large deviation from the trends that the data originally suggests. Regularization is a method of making a model that has become complicated due to overfitting into a simpler model, and is one of the highly versatile overfitting prevention methods that can be used in analysis methods such as regression and classification. The loss function is an index that indicates the error between the created model and the actual data, and the smaller the value, the better the model fits the data. In linear regression, for example, the mean squared error (MAE) is used.
[0018] The multiple regression analysis using regularization is explained below. The explanatory variables, predicted values, and regression coefficients are expressed as in the following formula (2). x is the explanatory variable, and w is the regression coefficient. The explanatory variables, predicted values, and regression coefficients are all expressed as vectors.
number
[0019] The objective variable can be expressed by the following formula (3) or the following formula (4). 0 x 0 represents the intercept.
number
number
[0020] When the above-mentioned objective variable is expressed as a vector, it becomes as shown in the following formula (5).
number
[0021] In multiple regression analysis, the explanatory variable x is not one, but a large number of training data are used to derive the regression coefficients. Therefore, there are as many predicted values as there are training data. If the sample size is n, it is expressed as in the following formula (6).
number
[0022] The regression coefficient w is determined so that there is no discrepancy between the predicted value and the actual value y. In multiple regression analysis, the least squares method is used to determine parameters so as to minimize the loss function L expressed by the following formula (7).
number
[0023] In regularization, a function obtained by adding a regularization term to the loss function L is treated as a new minimization target. Representative regularization terms include the following formula (8) or formula (9). λ is a hyperparameter whose value may be freely determined.
number
number
[0024] By minimizing a loss function with a regularization term added, it is possible to reduce the regression coefficients of variables that do not contribute to the objective variable, thereby reducing the number of effective explanatory variables and suppressing overfitting, compared to when a regularization term is not added.
[0025] Figure 1 shows an example of a two-dimensional explanatory variable. The horizontal axis is the explanatory variable x 1 Regression coefficient w 1 and the vertical axis is the explanatory variable x 2 Regression coefficient w 2 When the regularization term is zero, the loss function is optimized to minimize it based on the least squares method, and w = (w 1 ,w 2 ) is found. This unique solution is the point in the "Least Squares Solution" in Figure 1. When the regularization term is not zero, the regression coefficient w 1 and the regression coefficient w 2 changes from this point. If the loss function is the same value (on the contour line), in order to minimize the loss function L, which is the sum of the regularization term, "minimize the regularization term (w 1 ,w 2 )” or “the tangent to the contour of the regularization term” is the solution.
[0026] The contour of the regularization term in L2 regularization is shown in Figure 2 as an example: (w 1 ,w 2 ) On a plane, it is represented by a circle centered at the origin, so the solution that minimizes the loss function L is the tangent to the contour line of the loss function and the circle centered at the origin.
[0027] As the regularization term ratio λ is increased, a smaller regularization term contributes more to minimizing the loss function L, so the contours of the regularization term and loss function that create the solution move inward and outward, respectively. Here, in the example in Figure 2, the contours of the regularization term are perfect circles symmetrical on each axis, while the contours of the loss function L are small, with a small contribution from the regression coefficient w 1Due to this asymmetry, when λ is increased as shown in Figure 2, the solution, i.e., the position of the tangent line, changes with the regression coefficient w 1 The regression coefficient w 2 As a result, the contribution to the objective variable (≒ contribution to regression performance) of the regression coefficient w 1 Selectively reduce the regression coefficient w 2 It is possible to realize a situation where the change in
[0028] Regression coefficient w that has a small contribution to the objective variable 1 By reducing the number of regression coefficients, w 2 This prevents regression performance degradation without reducing the number of iterations, i.e., regularization can reduce overfitting while suppressing performance degradation.
[0029] For L1 regularization, the contour of the regularization term is (w 1 ,w 2 ) plane, it becomes a diamond (a square tilted at 45 degrees) as shown in Figure 3. The characteristic of L1 regularization is that the vertex of the contour line of the regularization term becomes the solution, and the regression coefficient w 1 becomes zero. In this case, one explanatory variable with a zero regression coefficient can be completely deleted, providing a stronger effect in preventing overfitting than L2 regularization.
[0030] In this embodiment, the spectral data obtained as the analysis result for the target substance is used, and the explanatory variables are the spectral data of each of the multiple elemental substances contained in the target substance. The analysis method may be, for example, Raman spectroscopy, infrared absorption spectroscopy, X-ray photoelectron spectroscopy, X-ray absorption spectroscopy, etc. The regression analysis is used to determine the combination ratio (composition ratio) of the spectra of the elements that reproduce (or fit) the spectral data of the target substance. In this case, the above-mentioned regularization is used to prevent overlearning, and to improve accuracy and efficiency. In addition, a threshold is set for the regression coefficient, and if there is a regression coefficient below the threshold, the elemental substance represented by the regression coefficient is excluded, and the regression analysis is repeated.
[0031] Specifically, the information in Table 1 is obtained from the spectral data obtained as the analysis result for the target substance and the spectral data for each of the multiple element substances contained in the target substance. Next, the obtained information is compared with the above formulas (2) to (9) and reduced to an optimization problem of the regression function shown in the following formula (10). The optimized β then becomes the composition ratio. In the following formulas (10) and (11), the first term on the right side is a loss function, and the second term is a regularization term. λ is a regularization parameter. The influence of the regularization term is adjusted using λ. In this embodiment, λ is scanned until β indicates a constant value, and the constant value is set as the optimal value, that is, the composition ratio. [Table 1]
number
number
[0032] "n" is the number of input samples (the number of x values on the horizontal axis of the spectrum). "p" is the number of explanatory variables (the number of spectra of individual components). The following formula (12) is the response variable (spectral intensity of compound / mixture) vector. The following formula (13) is the explanatory variable (spectral intensity of individual components) matrix. The following formula (14) is the regression coefficient vector.
number
number
number
[0033] The effects of this embodiment were verified by the following procedure.
[0034] (Step 1) Through simulation, infrared absorption spectra (IR spectra) were generated for the following four substances (substance numbers EQ0, EQ1, EQ2, and EQ3). EQ0: Pentaprismane EQ1: 1-Phenylcyclobutene EQ2: Vinyl-Cycolo-Octatetraene EQ3:Annulene
[0035] (Step 2) The four IR spectra generated in step 1 were synthesized by changing the composition ratio as follows (five combinations). Sample 1 EQ0:EQ1:EQ2:EQ3=0:60:30:10 Sample 2 EQ0:EQ1:EQ2:EQ3=0:30:60:10 Sample 3 EQ0:EQ1:EQ2:EQ3=0:10:20:70 Sample 4 EQ0:EQ1:EQ2:EQ3=0:0:20:80 Sample 5 EQ0:EQ1:EQ2:EQ3=0:0:10:90
[0036] (Step 3) The spectral data synthesized in step 2 was used as the objective variable, and the spectral data in step 1 was used as the explanatory variable, and the ratio of the spectrum used as the explanatory variable in the regression analysis using regularization according to this embodiment was specified to fit the spectrum of the objective variable. During regularization, the value of λ was scanned, and the value when each regression coefficient became constant or converged to a constant value was determined to be the final composition ratio.
[0037] Fig. 4(a) to Fig. 4(d) show IR spectrum data obtained by simulation. Fig. 4(a) shows IR spectrum data for EQ0. Fig. 4(b) shows IR spectrum data for EQ1. Fig. 4(c) shows IR spectrum data for EQ2. Fig. 4(d) shows IR spectrum data for EQ3. As shown in Table 2, the intensities of the spectrum data in Fig. 4(a) to Fig. 4(d) were used as explanatory variables, and the intensity of the spectrum data synthesized as described above was used as the objective variable. A regression analysis using regularization was performed according to the above formulas (10) and (11). In this case, λ was set to 1×10 2 From 1×10 -2 was scanned in descending order. [Table 2]
[0038] FIG. 5 is a graph showing the regression analysis. FIGS. 6(a) to 6(e) show the results of the regression analysis using the regularization according to this embodiment. FIG. 6(a) shows the results for Sample 1. FIG. 6(b) shows the results for Sample 2. FIG. 6(c) shows the results for Sample 3. FIG. 6(d) shows the results for Sample 4. FIG. 6(e) shows the results for Sample 5. If the optimum value is the value at which the coefficient β on the vertical axis becomes constant or converges to a constant value with respect to the change in λ on the horizontal axis, the initial synthesis ratio is accurately reproduced in both cases.
[0039] As described above, according to this embodiment, the spectral data of each of the plurality of component substances contained in the target substance is used as an explanatory variable, and the spectral data obtained as the analysis result for the target substance is used as a response variable to perform regularization, thereby identifying the composition ratio of the plurality of component substances, thereby simplifying a complicated process.
[0040] Next, a specific device configuration of this embodiment will be described. Fig. 7(a) is a functional block diagram illustrating an example of the overall configuration of an information processing device 100 according to the embodiment 1. As illustrated in Fig. 7(a), the information processing device 100 functions as an acquisition unit 10, a vector creation unit 20, a parameter setting unit 30, an estimation unit 40, a matrix creation unit 50, an analysis unit 60, an output unit 70, and the like.
[0041] Fig. 7(b) is a block diagram illustrating an example of a hardware configuration of the information processing device 100. As illustrated in Fig. 7(b), the information processing device 100 includes a CPU 101, a RAM 102, a storage device 103, an input device 104, a display device 105, and the like.
[0042] The CPU (Central Processing Unit) 101 is a central processing unit. The CPU 101 includes one or more cores. The RAM (Random Access Memory) 102 is a volatile memory that temporarily stores a program executed by the CPU 101, data processed by the CPU 101, and the like. The storage device 103 is a non-volatile storage device. For example, a ROM (Read Only Memory), a solid state drive (SSD) such as a flash memory, or a hard disk driven by a hard disk drive can be used as the storage device 103. The storage device 103 stores a calculation program. The input device 104 is an input device such as a keyboard or a mouse. The display device 105 is a display device such as an LCD (Liquid Crystal Display). The CPU 101 executes the calculation program to realize an acquisition unit 10, a vector creation unit 20, a parameter setting unit 30, an estimation unit 40, a matrix creation unit 50, an analysis unit 60, an output unit 70, and the like. The acquisition unit 10, the vector creation unit 20, the parameter setting unit 30, the estimation unit 40, the matrix creation unit 50, the analysis unit 60, the output unit 70, and the like may be implemented using hardware such as dedicated circuits.
[0043] 8 and 9 are flowcharts showing an example of the operation of the information processing device 100. As illustrated in Fig. 8 and Fig. 9, the acquisition unit 10 acquires spectral data as a result of an analysis of a target substance (step S1).
[0044] Next, the vector creation unit 20 creates an intensity vector of the above formula (12) for the spectrum acquired in step S1 (step S2).
[0045] In parallel with steps S1 to S2, the parameter setting unit 30 sets λmin and λmax (step S3). λmin is the minimum value of the parameter λ. λmax is the maximum value of the parameter λ. By setting λmin and λmax, the scanning range of the parameter λ can be set.
[0046] Next, the parameter setting unit 30 sets λt (step S4), which is the scanning step of the parameter λ and is the interval value of the parameter λ.
[0047] Next, the parameter setting unit 30 sets a threshold value th for the regression coefficient β (step S5).
[0048] Next, the parameter setting unit 30 sets the parameter λ to λmax (step S6).
[0049] In parallel with steps S1 to S2 and steps S3 to S6, the estimation unit 40 estimates each of the constituent substances of the target substance (step S7). In step S7, the estimation unit 40 may automatically estimate each of the constituent substances of the target substance, or may obtain each of the constituent substances input by the user using the input device 104.
[0050] Next, the matrix creation unit 50 acquires single-substance spectral data for each of the element substances obtained in step S7 (step S8). In step S8, the single-substance spectral data obtained by a user through a previous analysis or the like is acquired via the input device 104 or the like.
[0051] Next, the matrix creation unit 50 creates an intensity matrix of the above formula (13) for each of the spectral data of the simple substances acquired in step S8 (step S9).
[0052] After executing steps S2, S6, and S9, the analysis unit 60 performs regularized regression analysis according to the above formula (11) using the intensity vector created in step S2 as the objective variable and the intensity matrix created in step S9 as the explanatory variable (step S10).
[0053] Next, the analysis unit 60 determines whether the regression coefficients β of all the component substances indicate a constant value (step S11).
[0054] If the result of the determination in step S11 is "No," the analysis unit 60 sets λ to λ-λt (step S12), which makes it possible to reduce λ by the step value.
[0055] Next, the analysis unit 60 judges whether or not λ remains equal to or greater than λmin (step S13). By executing step S13, it is possible to judge whether or not the scanning of λ has been completed.
[0056] If the answer to step S13 is "Yes," the analysis unit 60 determines whether or not there is an element substance whose coefficient β is smaller than the threshold value th (step S14).
[0057] If the answer in step S14 is "Yes", the analysis unit 60 excludes element substances whose regression coefficient β is less than the threshold value th (step S15). After that, the process is executed again from step S10. If the answer in step S14 is "No", the process is executed again from step S10 without executing step S15.
[0058] If the result of the determination in step S13 is "No," steps S3 to S6 are executed again. In this case, at least one of λmin, λmax, λt, and the threshold value th is set to a value different from the previous value.
[0059] If the answer is "Yes" in step S11, the output unit 70 outputs the above formula (11) calculated in step S10 as the composition ratio (step S16). The output result of the output unit 70 is displayed on the display device 105 or the like. Thereafter, the execution of the flowchart ends.
[0060] In the above example, the analysis unit 60 is an example of an identification unit that performs regularization using the spectral data of each of the multiple component substances contained in the target substance as explanatory variables and the spectral data obtained as the analysis result for the target substance as a response variable, and identifies the composition ratio of the multiple component substances.
[0061] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to such specific embodiments, and various modifications and variations are possible within the scope of the gist of the present invention described in the claims. (Appendix 1) On the computer, a calculation program for executing a process of performing regularization using spectral data of each of a plurality of element substances contained in an analyte as explanatory variables and spectral data obtained as an analysis result for the analyte as a response variable, and identifying a composition ratio of the plurality of element substances in the analyte. (Appendix 2) The calculation program according to claim 1, wherein regression analysis is used when identifying the composition ratio. (Appendix 3) The computer includes: 3. The calculation program according to claim 2, further comprising: executing a process of identifying the composition ratio of the objective variable from a regression coefficient of the spectrum data of the simple substance in the regression analysis. (Appendix 4) The computer includes: 4. The arithmetic program according to claim 3, further comprising: a process of repeatedly executing the regression analysis by excluding an explanatory variable when the regression coefficient of the explanatory variable obtained as a result of the regression analysis is below a threshold value, until there are no regression coefficients below the threshold value. (Appendix 5) The computer includes: 3. The arithmetic program according to claim 2, characterized in that the arithmetic program executes a process of repeating the regression analysis while changing a parameter representing a weight of a regularization term used in the regularization. (Appendix 6) The plurality of component substances includes at least two or more kinds of compounds of the constituent components or a mixture of the two or more kinds of the constituent components; The calculation program according to any one of Supplementary Note 1 to Supplementary Note 4, characterized in that the spectral data of the single substances is spectral data of the two or more types of constituent components or predicted constituent components alone. (Appendix 7) The computer a calculation method comprising: performing a process of performing regularization using spectral data of each of a plurality of element substances contained in a target substance as explanatory variables and spectral data obtained as an analysis result for the target substance as a response variable, and identifying a composition ratio of the plurality of element substances in the target substance. (Appendix 8) The calculation method according to claim 7, wherein regression analysis is used when determining the composition ratio. (Appendix 9) The computer, The calculation method according to claim 8, characterized in that in the regression analysis, a process is performed in which the composition ratio is identified from a regression coefficient of the spectral data of the simple substance as the response variable. (Appendix 10) The computer, 10. The calculation method according to claim 9, wherein, when a regression coefficient of the explanatory variable obtained as a result of the regression analysis falls below a threshold, the explanatory variable is excluded and the regression analysis is performed, and the process is repeatedly performed until there are no regression coefficients below the threshold. (Appendix 11) The computer, The calculation method according to claim 8, characterized in that a process of repeating the regression analysis is executed while changing a parameter representing a weight of a regularization term used in the regularization. (Appendix 12) The plurality of component substances includes at least two or more kinds of compounds of the constituent components or a mixture of the two or more kinds of the constituent components; The calculation method according to any one of Supplementary Note 7 to Supplementary Note 11, characterized in that the spectral data of the single substance is spectral data of the two or more types of constituent components or predicted constituent components alone. (Appendix 13) 1. An information processing device comprising: an identification unit that uses spectral data of each of a plurality of component substances contained in an analyte as an explanatory variable, and performs regularization using spectral data obtained as an analysis result for the analyte as a target variable, and identifies a composition ratio of the plurality of component substances in the analyte. (Appendix 14) The information processing device according to claim 13, wherein the determination unit uses regression analysis when determining the composition ratio. (Appendix 15) The information processing device according to claim 14, wherein the identifying unit identifies the composition ratio from a regression coefficient of the spectral data of the simple substance as the objective variable in the regression analysis. (Appendix 16) The information processing device according to claim 15, wherein the identification unit repeatedly executes a process of excluding an explanatory variable and executing the regression analysis when a regression coefficient of the explanatory variable obtained as a result of the regression analysis is below a threshold value, until there is no regression coefficient below the threshold value. (Appendix 17) The information processing device according to claim 14, wherein the identification unit executes a process of repeating the regression analysis while changing a parameter representing a weight of a regularization term used in the regularization. (Appendix 18) The plurality of component substances includes at least two or more kinds of compounds of the constituent components or a mixture of the two or more kinds of the constituent components; The information processing device according to any one of claims 13 to 17, characterized in that the spectral data of the single entity is spectral data of the two or more types of constituent components or predicted constituent components alone. [Explanation of symbols]
[0062] 10 Acquisition Department 20 Vector Creation Section 30 Parameter setting section 40 Estimation part 50 Matrix Creation Section 60 Analysis Department 70 Output section 100 Information processing device 101 CPU 102 RAM 103 Storage device 104 Input Devices 105 Display device
Claims
1. On the computer, a calculation program for executing a process of performing regularization using spectral data of each of a plurality of element substances contained in an analyte as explanatory variables and spectral data obtained as an analysis result for the analyte as a response variable, and identifying a composition ratio of the plurality of element substances in the analyte.
2. 2. The calculation program according to claim 1, wherein a regression analysis is used when the composition ratio is specified.
3. The computer includes:
3. The computing program according to claim 2, further comprising a step of: determining, in the regression analysis, the composition ratio of the objective variable from a regression coefficient of the spectrum data of the simple substance.
4. The computer includes:
4. The computing program according to claim 3, further comprising a step of repeatedly executing a process of excluding an explanatory variable and executing the regression analysis when the regression coefficient of the explanatory variable obtained as a result of the regression analysis falls below a threshold value, until there are no regression coefficients below the threshold value.
5. The computer includes:
3. The computing program according to claim 2, further comprising a step of repeating the regression analysis while changing a parameter representing a weight of a regularization term used in the regularization.
6. The plurality of constituent substances includes at least two or more kinds of compounds of constituents or a mixture of the two or more kinds of constituents; 6. The computing program according to claim 1, wherein the spectral data of the single entity is spectral data of the two or more types of constituent components or predicted constituent components alone.
7. The computer a calculation method comprising: performing a process of performing regularization using spectral data of each of a plurality of element substances contained in a target substance as explanatory variables and spectral data obtained as an analysis result for the target substance as a response variable, and identifying a composition ratio of the plurality of element substances in the target substance.
8. 1. An information processing device comprising: an identification unit that uses spectral data of each of a plurality of element substances contained in an analyte as an explanatory variable, and performs regularization using spectral data obtained as an analysis result for the analyte as a target variable, and identifies a composition ratio of the plurality of element substances in the analyte.
Citation Information
Patent Citations
Ensemble method and apparatus for classifying materials and quantifying the components of mixtures
JP2010520471A
System, method, and apparatus for optical hydrocarbon gas composition monitoring
JP2019215356A
Component identification device for spectrum analyzer, method thereof, and computer program
JP2020201174A
Information processing device, method for controlling information processing device, and program
JP2021009135A
