Nervonic acid purification method and system based on multiple linear regression analysis

Abstract

The invention relates to the technical field of nervonic acid purification, in particular to a nervonic acid purification method and system based on multivariate linear regression analysis, and the method comprises the following steps: carrying out normality test on standard nervonic acid purity data to obtain a maximum correlation value, carrying out correlation analysis on the standard nervonic acid purity data by using a correlation analysis method to obtain a target correlation coefficient, carrying out descriptive statistics on the standard nervonic acid purity data to obtain a descriptive statistical result, preprocessing the verified nervonic acid purity data based on the descriptive statistical result and the target correlation coefficient to obtain target purity data, and carrying out model training on the initial purity model according to the target purity data to obtain a target purity model; and high-purity extraction of nervonic acid is realized based on the target purity model. The method can solve the problem of low nervonic acid extraction purity.

Description

Technical Field

[0001] The present invention relates to the technical field of neuraminic acid purification, and in particular to a neuraminic acid purification method and system based on multivariate linear regression analysis. Background Art

[0002] Multiple linear regression analysis is a statistical analysis method used to study the relationship between multiple independent variables and a continuous dependent variable. It aims to establish a linear model to examine the influence of multiple independent variables on the dependent variable and the correlation between them. The basic idea of ​​multiple linear regression analysis is to use linear equations to describe the relationship between independent variables and dependent variables, where the dependent variable is assumed to be a linear combination of multiple independent variables plus an error term, which represents random differences that cannot be explained by the model.

[0003] Currently, the purification of neuraminic acid is mainly achieved through solvent extraction, chromatographic separation, crystallization purification, reverse osmosis membrane and other technologies. However, there is a problem of low purity of neuraminic acid due to poor purification conditions. Therefore, the current method of purifying neuraminic acid through solvent extraction, chromatographic separation, crystallization purification, reverse osmosis membrane and other technologies has the problem of low purity of neuraminic acid extraction. Summary of the invention

[0004] The present invention provides a method and system for purifying neuraminic acid based on multivariate linear regression analysis, the main purpose of which is to solve the problem of low purity of neuraminic acid extraction.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for purifying neuraminic acid based on multivariate linear regression analysis, including: obtaining standard neuraminic acid purity data under different temperatures, pressures, solvent types and reaction times, performing a normality test on the standard neuraminic acid purity data to obtain a maximum correlation value; selecting a correlation analysis method based on the maximum correlation value, and performing a correlation analysis on the standard neuraminic acid purity data using the correlation analysis method to obtain a target correlation coefficient; performing descriptive statistics on the standard neuraminic acid purity data to obtain descriptive statistical results; obtaining verification neuraminic acid purity data, and preprocessing the verification neuraminic acid purity data based on the descriptive statistical results and the target correlation coefficient to obtain target purity data; establishing an initial purity model using a preset multivariate linear regression analysis method, and performing model training on the initial purity model according to the target purity data to obtain a target purity model; predicting the optimal purification process combination based on the target purity model, and achieving high-purity extraction of neuraminic acid according to the optimal purification process combination.

[0006] Optionally, the normality test of the standard neuraminic acid purity data to obtain the maximum correlation value includes: obtaining the number of neuraminic acid purity samples of the standard neuraminic acid purity data, and comparing the size relationship between the number of neuraminic acid purity samples and a preset sample threshold; if the number of neuraminic acid purity samples is less than or equal to the sample threshold, then using the preset Shapiro-Wilk test to perform a normality test on the standard neuraminic acid purity data to obtain the maximum correlation value; if the number of neuraminic acid purity samples is greater than the sample threshold, then using the preset Kolmogorov-Smirnov test to perform a normality test on the standard neuraminic acid purity data to obtain the maximum correlation value.

[0007] Optionally, the correlation analysis method is selected based on the maximum correlation value, and the correlation analysis method is used to perform correlation analysis on the standard neuraminic acid purity data to obtain a target correlation coefficient, including: determining whether the maximum correlation value is greater than 0.05; if the maximum correlation value is greater than 0.05, calculating the Pearson correlation coefficient of the standard neuraminic acid purity data, and using the Pearson correlation coefficient as the target correlation coefficient; if the maximum correlation is not greater than 0.05, calculating the Spearman correlation coefficient of the standard neuraminic acid purity data, and using the Spearman correlation coefficient as the target correlation coefficient.

[0008] Optionally, the standard neuraminic acid purity data is subjected to descriptive statistics to obtain descriptive statistical results, including: identifying the mean and median of the standard neuraminic acid purity data, and calculating the central trend difference of the standard neuraminic acid purity data based on the mean and median; calculating the median absolute deviation of the neuraminic acid purity data using the median; drawing a box plot based on the standard neuraminic acid purity data; and combining the mean, central trend difference, median absolute deviation and box plot to obtain descriptive statistical results.

[0009] Optionally, the verification neuraminic acid purity data is preprocessed based on the descriptive statistical results and the target correlation coefficient to obtain the target purity data, including: extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; calculating the verification correlation coefficient and the descriptive verification result according to the standard neuraminic acid purity data and the verification neuraminic acid purity parameter set; respectively calculating the correlation error between the verification correlation coefficient and the target correlation coefficient and the statistical error between the descriptive verification result and the descriptive statistical result; calculating the comprehensive error according to the correlation error and the statistical error; judging whether the comprehensive error is greater than a preset error threshold; if the comprehensive error is greater than the error threshold, returning to the above-mentioned The step of extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; if the comprehensive error is not greater than the error threshold, merging the verification neuraminic acid purity parameter set into the standard neuraminic acid purity data to obtain merged neuraminic acid purity data; judging whether the verification neuraminic acid purity data has completed the extraction of the verification neuraminic acid purity parameter set; if the verification neuraminic acid purity data has not completed the extraction of the verification neuraminic acid purity parameter set, returning to the above-mentioned step of extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; if the verification neuraminic acid purity data has completed the extraction of the verification neuraminic acid purity parameter set, taking the merged neuraminic acid purity data as the target purity data.

[0010] Optionally, the initial purity model is established by using a preset multiple linear regression analysis method, including: taking temperature, pressure, solvent type, and reaction time as independent variables, and taking a preset predicted nervonic acid purity as a dependent variable; according to the independent variables and the dependent variables, the initial purity model is established by using a multiple linear regression analysis method, wherein the initial purity model is: ,in, Indicates the predicted purity of nervonic acid, represents the intercept, represents the temperature regression coefficient, Indicates temperature, represents the pressure regression coefficient, Indicates pressure, represents the solvent regression coefficient, Indicates the solvent type, represents the time regression coefficient, Indicates reaction time.

[0011] Optionally, the initial purity model is trained according to the target purity data to obtain the target purity model, including: dividing the target purity data into k non-overlapping sub-datasets; extracting sub-datasets in the k non-overlapping sub-datasets in turn, using the sub-datasets as validation sets, and removing the validation sets from the k non-overlapping sub-datasets to obtain a training set; substituting the training set into the initial purity model to obtain a training neuraminic acid purity set; calculating the regression sum of squares and the total sum of squares using the training neuraminic acid purity set and the training set; calculating the determination coefficient based on the regression sum of squares and the total sum of squares; judging whether the determination coefficient is less than a preset prediction error threshold; if the determination coefficient is not less than the prediction error threshold , then use the determination coefficient to adjust the parameters of the initial purity model to obtain an iteratively adjusted purity model; use the iteratively adjusted purity model to update the initial purity model, and return to the above-mentioned step of sequentially extracting sub-datasets from the k non-overlapping sub-datasets; if the determination coefficient is less than the prediction error threshold, calculate the mean square error of the validation set according to the iteratively adjusted purity model; determine whether the mean square error is less than the preset mean square error threshold; if the mean square error is not less than the mean square error threshold, return to the above-mentioned step of sequentially extracting sub-datasets from the k non-overlapping sub-datasets; if the mean square error is less than the mean square error threshold, use the iteratively adjusted purity model as the target purity model.

[0012] Optionally, the method of using the determined coefficient to adjust the parameters of the initial purity model to obtain an iteratively adjusted purity model includes: extracting initial coefficients preset in the initial purity model in sequence; setting a learning rate according to the determined coefficient, and adjusting the initial coefficient using the learning rate to obtain an adjusted coefficient; and updating the initial coefficient using the adjusted coefficient to obtain an iteratively adjusted purity model.

[0013] Optionally, predicting the optimal purification process combination based on the target purity model includes: obtaining a purification process combination set, and extracting purification process combinations in the purification process combination set in sequence; inputting the purification process combination into the target purity model to obtain a predicted product purity value set; extracting the maximum product purity value from the predicted product purity value set, and identifying the optimal purification process combination corresponding to the maximum product purity value.

[0014] To achieve the above-mentioned objectives, the present invention also provides a neuraminic acid purification system based on multivariate linear regression analysis, including: a correlation analysis and descriptive statistics module, which is used to obtain standard neuraminic acid purity data under different temperatures, pressures, solvent types and reaction times, and perform a normality test on the standard neuraminic acid purity data to obtain a maximum correlation value; select a correlation analysis method based on the maximum correlation value, and perform a correlation analysis on the standard neuraminic acid purity data using the correlation analysis method to obtain a target correlation coefficient; perform descriptive statistics on the standard neuraminic acid purity data to obtain a descriptive statistical result; a neuraminic acid purity data preprocessing module, which is used to obtain neuraminic acid purity data, and preprocess the neuraminic acid purity data based on the descriptive statistical result and the target correlation coefficient to obtain the target purity data; an initial purity model training module, which is used to establish an initial purity model using a preset multivariate linear regression analysis method, and perform model training on the initial purity model according to the target purity data to obtain a target purity model; an optimal purification process combination prediction module, which is used to predict the optimal purification process combination based on the target purity model, and achieve high-purity extraction of neuraminic acid according to the optimal purification process combination.

[0015] The present invention is to solve the problem described in the background technology. The present invention obtains the standard neuraminic acid purity data under different temperatures, pressures, solvent types and reaction times, performs normality test, descriptive statistics and correlation analysis on the standard neuraminic acid purity data, obtains the maximum correlation value, descriptive statistics results and target correlation coefficient, and the standard neuraminic acid purity data has the problem of small data volume, so it is necessary to use the verification neuraminic acid purity data to supplement the data, and in the process of supplementing the data, the verification neuraminic acid purity data can be pre-processed based on the descriptive statistics results and the target correlation coefficient, and finally the target purity data is obtained, and the initial purity model is established by using the preset multiple linear regression analysis method, and the initial purity model can be trained by the target purity data to obtain the target purity model, and finally the optimal purification process combination is predicted based on the target purity model to achieve high-purity extraction of neuraminic acid. Therefore, the present invention can solve the problem that the current neuraminic acid extraction purity is not high. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a method for purifying neuraminic acid based on multivariate linear regression analysis provided in one embodiment of the present invention.

[0017] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0019] The embodiment of the present application provides a method for purifying neuraminic acid based on multiple linear regression analysis. The execution subject of the method for purifying neuraminic acid based on multiple linear regression analysis includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the method for purifying neuraminic acid based on multiple linear regression analysis can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0020] Reference Figure 1 As shown, it is a schematic diagram of a process of nervonic acid purification method based on multiple linear regression analysis provided by one embodiment of the present invention. In this embodiment, the nervonic acid purification method based on multiple linear regression analysis includes: S1, obtaining standard nervonic acid purity data under different temperatures, pressures, solvent types and reaction times, performing a normality test on the standard nervonic acid purity data, and obtaining a maximum correlation value.

[0021] It can be explained that the solvent type refers to the type of purification solvent used in the purification process of nervonic acid, such as n-hexane, ethyl acetate, petroleum ether, acetone, methanol, ethanol and (supercritical fluid solvent) and other solvents. The temperature, pressure and reaction time refer to the ultrasonic temperature, ultrasonic pressure and ultrasonic reaction time during the purification of neuraminic acid, respectively. The standard neuraminic acid purity data refers to the standard data on the correspondence between different ultrasonic temperatures, ultrasonic pressures, solvent types and ultrasonic reaction times and neuraminic acid purity. The standard data is neuraminic acid purification data that has been manually screened and does not contain abnormal, missing, erroneous values, for example: when the ultrasonic temperature is 90 When the ultrasonic pressure is 27 MPa, the solvent type is ethyl acetate, and the ultrasonic reaction time is 6 h, the purity of the nervonic acid can be 23%; when the ultrasonic temperature is 79 , when the ultrasonic pressure is 25 MPa, the solvent type is ethyl acetate, and the ultrasonic reaction time is 7 hours, the purity of the neuraminic acid can be 25%, and so on.

[0022] Furthermore, the normality test refers to testing whether the standard neuraminic acid purity data conforms to the normal distribution, and the maximum correlation value refers to a numerical value for determining whether the standard neuraminic acid purity data conforms to the normal distribution.

[0023] In an embodiment of the present invention, the normality test of the standard neuraminic acid purity data to obtain the maximum correlation value includes: obtaining the number of neuraminic acid purity samples of the standard neuraminic acid purity data, and comparing the size relationship between the number of neuraminic acid purity samples and a preset sample threshold; if the number of neuraminic acid purity samples is less than or equal to the sample threshold, using the preset Shapiro-Wilk test to perform a normality test on the standard neuraminic acid purity data to obtain the maximum correlation value; if the number of neuraminic acid purity samples is greater than the sample threshold, using the preset Kolmogorov-Smirnov test to perform a normality test on the standard neuraminic acid purity data to obtain the maximum correlation value.

[0024] It can be understood that the number of neuraminic acid purity samples refers to the number of samples in the standard neuraminic acid purity data, the Shapiro-Wilk test is applicable to small sample data, and the Kolmogorov-Smirnov test is applicable to large sample data, so the sample threshold can be used to judge the sample data size of the standard neuraminic acid purity data. When the number of neuraminic acid purity samples is greater than the sample threshold, it indicates that the standard neuraminic acid purity data is large sample data, and the Kolmogorov-Smirnov test can be used to perform a normality test on the standard neuraminic acid purity data. When the number of neuraminic acid purity samples is less than or equal to the sample threshold, it indicates that the standard neuraminic acid purity data is small sample data, and the Shapiro-Wilk test can be used to perform a normality test on the standard neuraminic acid purity data. The sample threshold can be 50. The Shapiro-Wilk test and the Kolmogorov-Smirnov test are prior arts and will not be described in detail here.

[0025] It can be understood that in order to perform a normality test, the standard neuraminic acid purity data can be serially arranged according to the ultrasonic temperature, ultrasonic pressure, solvent type and ultrasonic reaction time to obtain a serialized permutation combination. The solvent type can be represented by numbers, for example: n-hexane is represented as 1, ethyl acetate is represented as 2, petroleum ether is represented as 3, and so on. When the ultrasonic reaction time interval is [6h, 8h], the serialized permutation combination can be 80 , 20Mpa, 1, 6h-20%; 80 , 20Mpa, 1, 6.1h-21%; 80 , 20Mpa, 1, 6.3h-22%; 80 , 20Mpa, 1, 6.4h-23%; 80 , 20Mpa, 1, 6.5h-24%; 80 , 20Mpa, 1, 6.9h-25%, etc., the corresponding permutation and combination numbers are 1, 2, 3, 4, 5, 6, etc. , 20Mpa, 1, 6h-20% means the ultrasonic temperature is 80 , the ultrasonic pressure is 20Mpa, the solvent type is n-hexane, the ultrasonic reaction time is 6h, and the purity of nervonic acid is 20%. The arrangement and combination can be determined in the form of a dendrogram.

[0026] S2. Select the correlation analysis method based on the maximum correlation value, and use the correlation analysis method to perform correlation analysis on the standard neuraminic acid purity data to obtain the target correlation coefficient.

[0027] It can be understood that the correlation analysis method refers to an analysis method for analyzing the degree of correlation between the purity of nervonic acid and the sequence number of the arrangement and combination. The target correlation coefficient refers to the coefficient of the correlation between the purity of nervonic acid and the sequence number of the arrangement and combination.

[0028] In an embodiment of the present invention, the correlation analysis method is selected based on the maximum correlation value, and the correlation analysis method is used to perform correlation analysis on the standard neuraminic acid purity data to obtain a target correlation coefficient, including: judging whether the maximum correlation value is greater than 0.05; if the maximum correlation value is greater than 0.05, calculating the Pearson correlation coefficient of the standard neuraminic acid purity data, and using the Pearson correlation coefficient as the target correlation coefficient; if the maximum correlation is not greater than 0.05, calculating the Spearman correlation coefficient of the standard neuraminic acid purity data, and using the Spearman correlation coefficient as the target correlation coefficient.

[0029] It can be understood that when the maximum correlation value is greater than the significance level (0.05), it means that the standard neuraminic acid purity data conforms to the normal distribution, and when the maximum correlation value is not greater than 0.05, it means that the standard neuraminic acid purity data does not conform to the normal distribution.

[0030] Furthermore, the calculation process of the Pearson correlation coefficient requires that the standard neuraminic acid purity data meets the prerequisites such as linear relationship or normal distribution, so the Pearson correlation coefficient is suitable for processing the standard neuraminic acid purity data under normal distribution. The calculation process of the Spearman correlation coefficient does not make assumptions about the distribution of the standard neuraminic acid purity data, so the Spearman correlation coefficient is applicable to various types of standard neuraminic acid purity data. Therefore, when the maximum correlation value is greater than 0.05 (i.e., the standard neuraminic acid purity data conforms to the normal distribution), the Pearson correlation coefficient of the standard neuraminic acid purity data can be calculated, and the Pearson correlation coefficient is used as the target correlation coefficient; when the maximum correlation value is not greater than 0.05 (i.e., the standard neuraminic acid purity data does not conform to the normal distribution), the Spearman correlation coefficient of the standard neuraminic acid purity data can be calculated, and the Spearman correlation coefficient is used as the target correlation coefficient.

[0031] Furthermore, the calculation formula of the Pearson correlation coefficient is as follows: ,in, represents the Pearson correlation coefficient, represents the jth purity of nervonic acid in the standard nervonic acid purity data, Represents the mean value of nervonic acid purity in the standard nervonic acid purity data, Indicates the jth permutation combination number in the standard neuraminic acid purity data, Represents the mean of the permutation combination numbers in the standard neuraminic acid purity data.

[0032] Furthermore, the calculation formulas for the Pearson correlation coefficient and the Spearman correlation coefficient are prior art and will not be described in detail here.

[0033] S3. Perform descriptive statistics on the purity data of standard neuraminic acid to obtain descriptive statistical results.

[0034] It can be understood that the descriptive statistics refer to the statistics of the basic characteristics of the data of the standard neuraminic acid purity data, and the basic characteristics of the data include: mean, central trend difference, median absolute deviation, box plot. The descriptive statistical results refer to the statistical results of the basic characteristics of the data of the standard neuraminic acid purity data.

[0035] In an embodiment of the present invention, descriptive statistics are performed on the standard neuraminic acid purity data to obtain descriptive statistical results, including: identifying the mean and median of the standard neuraminic acid purity data, and calculating the central trend difference of the standard neuraminic acid purity data based on the mean and median; using the median to calculate the median absolute deviation of the neuraminic acid purity data; drawing a box plot based on the standard neuraminic acid purity data; and combining the mean, central trend difference, median absolute deviation and box plot to obtain descriptive statistical results.

[0036] It can be understood that the central trend difference refers to the difference between the mean and the median. When the mean is greater than the median, it means that the standard neuraminic acid purity data is more right-skewed. When the mean is less than the median, it means that the standard neuraminic acid purity data is more left-skewed. The mean refers to the mean of neuraminic acid purity, and the median refers to the median of neuraminic acid purity.

[0037] Furthermore, the median absolute deviation refers to the absolute deviation of the purity of nervonic acid, and the calculation formula is as follows: ,in, represents the median absolute deviation, Refers to the median value function, represents the purity of the i-th nervonic acid in the standard nervonic acid purity data, Represents the median of the purity data of standard neuraminic acid, The symbol for absolute value. The median value function refers to a function for extracting the median from the difference between all nervonic acid purities and the median.

[0038] It is understandable that the box-plot, also known as a box-whisker plot or a box-type plot, is a statistical diagram that describes the dispersion of standard neuraminic acid purity data. The box-plot is a prior art and will not be described in detail here.

[0039] S4. Obtain the verification data of neuraminic acid purity, pre-process the verification data of neuraminic acid purity based on the descriptive statistical results and the target correlation coefficient, and obtain the target purity data.

[0040] Furthermore, the verification neuraminic acid purity data refers to neuraminic acid purification data used to expand the standard neuraminic acid purity data and may contain abnormal, missing, erroneous values, and the target purity data refers to the neuraminic acid purity data after the standard neuraminic acid purity data is supplemented by the verification neuraminic acid purity data.

[0041] In an embodiment of the present invention, the verification neuraminic acid purity data is preprocessed based on the descriptive statistical results and the target correlation coefficient to obtain the target purity data, including: extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; calculating the verification correlation coefficient and the descriptive verification result according to the standard neuraminic acid purity data and the verification neuraminic acid purity parameter set; respectively calculating the correlation error between the verification correlation coefficient and the target correlation coefficient and the statistical error between the descriptive verification result and the descriptive statistical result; calculating the comprehensive error according to the correlation error and the statistical error; judging whether the comprehensive error is greater than a preset error threshold; if the comprehensive error is greater than the error threshold, returning to the above step. The steps of extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; if the comprehensive error is not greater than the error threshold, merging the verification neuraminic acid purity parameter set into the standard neuraminic acid purity data to obtain merged neuraminic acid purity data; judging whether the verification neuraminic acid purity data has completed the extraction of the verification neuraminic acid purity parameter set; if the verification neuraminic acid purity data has not completed the extraction of the verification neuraminic acid purity parameter set, returning to the above-mentioned steps of extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; if the verification neuraminic acid purity data has completed the extraction of the verification neuraminic acid purity parameter set, taking the merged neuraminic acid purity data as the target purity data.

[0042] It can be understood that the parameter set for verifying the purity of neuraminic acid refers to the data of a single purification test in the data for verifying the purity of neuraminic acid, for example: , 20Mpa, 1, 6h-20%. The verification correlation coefficient refers to the target correlation coefficient of the overall data after the verification neuraminic acid purity parameter set is incorporated into the standard neuraminic acid purity data, the descriptive verification result refers to the descriptive statistical result of the overall data after the verification neuraminic acid purity parameter set is incorporated into the standard neuraminic acid purity data, the correlation error refers to the absolute value of the difference between the verification correlation coefficient and the target correlation coefficient, and the statistical error refers to the sum of the absolute values ​​of the difference between the descriptive verification result and the corresponding data features of the descriptive statistical result, such as: the absolute value of the difference in the mean, the absolute value of the difference in the central trend difference, the absolute value of the difference in the median absolute deviation, and the absolute value of the difference in the quartiles in the box plot. The comprehensive error refers to the weighted sum of the correlation error and the statistical error, and the weight of the correlation error and the weight of the statistical error can be set as needed.

[0043] Further, when the comprehensive error is not greater than the error threshold, it indicates that the verification neuraminic acid purity parameter set meets the data standard of the standard neuraminic acid purity data, and therefore, the verification neuraminic acid purity parameter set can be incorporated into the standard neuraminic acid purity data to obtain the merged neuraminic acid purity data after the standard neuraminic acid purity data is amplified. If the comprehensive error is greater than the error threshold, it indicates that the verification neuraminic acid purity parameter set does not meet the data standard of the standard neuraminic acid purity data, and therefore, the verification neuraminic acid purity parameter set can be directly eliminated.

[0044] S5. Use a preset multiple linear regression analysis method to establish an initial purity model, and perform model training on the initial purity model according to the target purity data to obtain a target purity model.

[0045] Explainably, the multivariate linear regression analysis method refers to a method for studying the linear relationship between the purity of nervonic acid (a dependent variable) and temperature, pressure, solvent type, and reaction time (multiple independent variables). The target purity model refers to the initial purity model after model training. The initial purity model is described in the following embodiments.

[0046] In an embodiment of the present invention, the initial purity model is established by using a preset multiple linear regression analysis method, including: taking temperature, pressure, solvent type, and reaction time as independent variables, and taking a preset predicted purity of nervonic acid as a dependent variable; according to the independent variables and the dependent variables, the initial purity model is established by using a multiple linear regression analysis method, wherein the initial purity model is: ,in, Indicates the predicted purity of nervonic acid, represents the intercept, represents the temperature regression coefficient, Indicates temperature, represents the pressure regression coefficient, Indicates pressure, represents the solvent regression coefficient, Indicates the solvent type, represents the time regression coefficient, Indicates reaction time.

[0047] In an embodiment of the present invention, the initial purity model is trained according to the target purity data to obtain the target purity model, including: dividing the target purity data into k non-overlapping sub-datasets; extracting sub-datasets in the k non-overlapping sub-datasets in turn, using the sub-datasets as validation sets, and removing the validation sets from the k non-overlapping sub-datasets to obtain a training set; substituting the training set into the initial purity model to obtain a training neuraminic acid purity set; calculating the regression sum of squares and the total sum of squares using the training neuraminic acid purity set and the training set; calculating the determination coefficient based on the regression sum of squares and the total sum of squares; judging whether the determination coefficient is less than a preset prediction error threshold; if the determination coefficient is not less than the prediction error threshold, use the determination coefficient to adjust the parameters of the initial purity model to obtain an iteratively adjusted purity model; use the iteratively adjusted purity model to update the initial purity model, and return to the above step of sequentially extracting sub-datasets from the k non-overlapping sub-datasets; if the determination coefficient is less than the prediction error threshold, calculate the mean square error of the validation set according to the iteratively adjusted purity model; determine whether the mean square error is less than the preset mean square error threshold; if the mean square error is not less than the mean square error threshold, return to the above step of sequentially extracting sub-datasets from the k non-overlapping sub-datasets; if the mean square error is less than the mean square error threshold, use the iteratively adjusted purity model as the target purity model.

[0048] Explainable, the non-overlapping sub-datasets refer to data sets without data duplication. The method for dividing the target purity data can refer to the k-fold cross-validation method. The training nervonic acid purity set refers to the predicted nervonic acid purity set obtained after multiple independent variables of each test in the training set are input into the initial purity model. For example: when the training set is 80 , 20Mpa, 1, 6h-20%; 80 , 20Mpa, 1, 6.1h-21%; 80 , 20Mpa, 1, 6.3h-22%; 80 , 20Mpa, 1, 6.4h-23%; 80 , 20Mpa, 1, 6.5h-24%; 80 , 20Mpa, 1, 6.9h-25%, the multiple independent variables of the tests are: 80 , 20Mpa, 1, 6h; 80 , 20Mpa, 1, 6.1h; 80 , 20Mpa, 1, 6.3h; 80 , 20Mpa, 1, 6.4h; 80 , 20Mpa, 1, 6.5h; 80 , 20Mpa, 1, 6.9h, after inputting the above multiple independent variables into the initial purity model in turn, the obtained training neuraminic acid purity set may be: 18%, 21%, 23%, 25%, 27%, 28%. The regression sum of squares refers to the sum of the squares of the difference between the dependent variable regression value (the predicted neuraminic acid purity in the training neuraminic acid purity set) and the dependent variable mean value (the mean of the actual neuraminic acid purity in the training set), and the total sum of squares refers to the sum of the square differences between the actual neuraminic acid purity in the training set and the mean of the actual neuraminic acid purity. For example, when the training neuraminic acid purity set is 18%, 21%, 23%, and 25%, and the corresponding actual neuraminic acid purity set in the training set is 20%, 21%, 22%, and 23%, then the mean of the actual neuraminic acid purity in the training set is 21.5%, and the regression sum of squares is , the total sum of squares is .

[0049] Furthermore, the determination coefficient refers to the absolute value of the difference between the regression sum of squares and the total sum of squares. The larger the determination coefficient, the worse the prediction effect of the initial purity model. The smaller the determination coefficient, the better the prediction effect of the initial purity model. The mean square error refers to the average value of the square of the difference between the actual purity of neuraminic acid in the validation set and the corresponding predicted purity of neuraminic acid. For example, when the actual purity of neuraminic acid in the validation set is 18%, 21%, 23%, and 25%, and the predicted purity of neuraminic acid is 19%, 20%, 22%, and 24%, the mean square error is .

[0050] In an embodiment of the present invention, the method of using the determined coefficient to adjust the parameters of the initial purity model to obtain an iteratively adjusted purity model includes: extracting the initial coefficients preset in the initial purity model in sequence; setting a learning rate according to the determined coefficient, adjusting the initial coefficient using the learning rate to obtain an adjusted coefficient; and updating the initial coefficient using the adjusted coefficient to obtain an iteratively adjusted purity model.

[0051] It is understandable that the initial coefficient refers to each coefficient in the initial purity model, such as intercept, temperature regression coefficient, pressure regression coefficient, solvent regression coefficient, and time regression coefficient. The learning rate refers to the speed and direction of updating the initial coefficient in the initial purity model during the parameter adjustment process. If the learning rate is set too large, the optimal solution may be crossed during each parameter adjustment process. If the learning rate is too small, the speed at which the initial coefficient converges to the optimal solution will be very slow. Therefore, when the determination coefficient is large, the learning rate should also increase accordingly, and when the determination coefficient is small, the learning rate should also decrease accordingly.

[0052] Furthermore, the adjustment process of the initial coefficients may be adjusted according to the gradient descent algorithm in combination with a learning rate. The learning rate may be a key hyperparameter in the gradient descent algorithm for controlling the step size of each initial coefficient update.

[0053] S6. Predict the optimal purification process combination based on the target purity model, and achieve high-purity extraction of neuraminic acid according to the optimal purification process combination.

[0054] Furthermore, the optimal purification process combination refers to the optimal independent variable combination (i.e., the optimal temperature, pressure, solvent type, and reaction time combination) predicted based on the target purity model.

[0055] In an embodiment of the present invention, predicting the best purification process combination based on the target purity model includes: obtaining a purification process combination set, and extracting purification process combinations in the purification process combination set in sequence; inputting the purification process combination into the target purity model to obtain a predicted product purity value set; extracting the maximum product purity value in the predicted product purity value set, and identifying the best purification process combination corresponding to the maximum product purity value.

[0056] It can be explained that the predicted product purity value set refers to the set of neuraminic acid purities calculated after the purification process combination set is input into the target purity model.

[0057] The present invention is to solve the problem described in the background technology. The present invention obtains the standard neuraminic acid purity data under different temperatures, pressures, solvent types and reaction times, performs normality test, descriptive statistics and correlation analysis on the standard neuraminic acid purity data, obtains the maximum correlation value, descriptive statistics results and target correlation coefficient, and the standard neuraminic acid purity data has the problem of small data volume, so it is necessary to use the verification neuraminic acid purity data to supplement the data, and in the process of supplementing the data, the verification neuraminic acid purity data can be pre-processed based on the descriptive statistics results and the target correlation coefficient, and finally the target purity data is obtained, and the initial purity model is established by using the preset multiple linear regression analysis method, and the initial purity model can be trained by the target purity data to obtain the target purity model, and finally the optimal purification process combination is predicted based on the target purity model to achieve high-purity extraction of neuraminic acid. Therefore, the present invention can solve the problem that the current neuraminic acid extraction purity is not high.

[0058] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for purifying nervonic acid based on multiple linear regression analysis, characterized in that: The method includes: obtaining standard neuraminic acid purity data under different temperatures, pressures, solvent types and reaction times, performing a normality test on the standard neuraminic acid purity data to obtain a maximum correlation value; selecting a correlation analysis method based on the maximum correlation value, performing a correlation analysis on the standard neuraminic acid purity data using the correlation analysis method to obtain a target correlation coefficient; performing descriptive statistics on the standard neuraminic acid purity data to obtain a descriptive statistical result; obtaining verification neuraminic acid purity data, preprocessing the verification neuraminic acid purity data based on the descriptive statistical result and the target correlation coefficient to obtain a target purity data; establishing an initial purity model using a preset multivariate linear regression analysis method, performing model training on the initial purity model according to the target purity data to obtain a target purity model; predicting an optimal purification process combination based on the target purity model, and achieving high-purity extraction of neuraminic acid according to the optimal purification process combination.

2. The method for purifying neuraminic acid based on multiple linear regression analysis as claimed in claim 1, characterized in that: The method of performing a normality test on the standard neuraminic acid purity data to obtain a maximum correlation value includes: obtaining the number of neuraminic acid purity samples of the standard neuraminic acid purity data, and comparing the size relationship between the number of neuraminic acid purity samples and a preset sample threshold; if the number of neuraminic acid purity samples is less than or equal to the sample threshold, performing a normality test on the standard neuraminic acid purity data using a preset Shapiro-Wilk test to obtain a maximum correlation value; if the number of neuraminic acid purity samples is greater than the sample threshold, performing a normality test on the standard neuraminic acid purity data using a preset Kolmogorov-Smirnov test to obtain a maximum correlation value.

3. The method for purifying neuraminic acid based on multiple linear regression analysis as claimed in claim 2, characterized in that: The correlation analysis method is selected based on the maximum correlation value, and the correlation analysis method is used to perform correlation analysis on the standard neuraminic acid purity data to obtain a target correlation coefficient, including: determining whether the maximum correlation value is greater than 0.05; if the maximum correlation value is greater than 0.05, calculating the Pearson correlation coefficient of the standard neuraminic acid purity data, and using the Pearson correlation coefficient as the target correlation coefficient; if the maximum correlation is not greater than 0.05, calculating the Spearman correlation coefficient of the standard neuraminic acid purity data, and using the Spearman correlation coefficient as the target correlation coefficient.

4. The method for purifying neuraminic acid based on multiple linear regression analysis as claimed in claim 3, characterized in that: The descriptive statistics of the standard neuraminic acid purity data are performed to obtain descriptive statistical results, including: identifying the mean and median of the standard neuraminic acid purity data, and calculating the central trend difference of the standard neuraminic acid purity data based on the mean and median; using the median to calculate the median absolute deviation of the neuraminic acid purity data; drawing a box plot based on the standard neuraminic acid purity data; combining the mean, central trend difference, median absolute deviation and box plot to obtain descriptive statistical results.

5. The method for purifying neuraminic acid based on multiple linear regression analysis as claimed in claim 4, characterized in that: The method of preprocessing the verification neuraminic acid purity data based on the descriptive statistical results and the target correlation coefficient to obtain the target purity data includes: extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; calculating the verification correlation coefficient and the descriptive verification result according to the standard neuraminic acid purity data and the verification neuraminic acid purity parameter set; respectively calculating the correlation error between the verification correlation coefficient and the target correlation coefficient and the statistical error between the descriptive verification result and the descriptive statistical result; calculating the comprehensive error according to the correlation error and the statistical error; judging whether the comprehensive error is greater than a preset error threshold; if the comprehensive error is greater than the error threshold, returning to the above-mentioned The steps of extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; if the comprehensive error is not greater than the error threshold, merging the verification neuraminic acid purity parameter set into the standard neuraminic acid purity data to obtain merged neuraminic acid purity data; judging whether the verification neuraminic acid purity data has completed the extraction of the verification neuraminic acid purity parameter set; if the verification neuraminic acid purity data has not completed the extraction of the verification neuraminic acid purity parameter set, returning to the above-mentioned steps of extracting the verification neuraminic acid purity parameter set in sequence from the verification neuraminic acid purity data; if the verification neuraminic acid purity data has completed the extraction of the verification neuraminic acid purity parameter set, taking the merged neuraminic acid purity data as the target purity data.

6. The method for purifying neuraminic acid based on multiple linear regression analysis according to claim 5, characterized in that: The method of establishing an initial purity model using a preset multiple linear regression analysis method includes: using temperature, pressure, solvent type, and reaction time as independent variables, and using a preset predicted purity of nervonic acid as a dependent variable; and establishing an initial purity model using a multiple linear regression analysis method based on the independent variables and the dependent variables, wherein the initial purity model is: ,in, Indicates the predicted purity of nervonic acid, represents the intercept, represents the temperature regression coefficient, Indicates temperature, represents the pressure regression coefficient, Indicates pressure, represents the solvent regression coefficient, Indicates the solvent type, represents the time regression coefficient, Indicates reaction time.

7. The method for purifying neuraminic acid based on multiple linear regression analysis according to claim 6, characterized in that: The method of training the initial purity model according to the target purity data to obtain the target purity model includes: dividing the target purity data into k non-overlapping sub-datasets; extracting sub-datasets in the k non-overlapping sub-datasets in turn, using the sub-datasets as validation sets, and removing the validation sets from the k non-overlapping sub-datasets to obtain a training set; substituting the training set into the initial purity model to obtain a training neuraminic acid purity set; calculating the regression sum of squares and the total sum of squares using the training neuraminic acid purity set and the training set; calculating the determination coefficient based on the regression sum of squares and the total sum of squares; judging whether the determination coefficient is less than a preset prediction error threshold; if the determination coefficient is not less than the prediction error threshold, then The determination coefficient is used to adjust the parameters of the initial purity model to obtain an iteratively adjusted purity model; the iteratively adjusted purity model is used to update the initial purity model, and the step of sequentially extracting sub-datasets from the k non-overlapping sub-datasets is returned to; if the determination coefficient is less than the prediction error threshold, the mean square error of the validation set is calculated according to the iteratively adjusted purity model; it is determined whether the mean square error is less than a preset mean square error threshold; if the mean square error is not less than the mean square error threshold, the step of sequentially extracting sub-datasets from the k non-overlapping sub-datasets is returned to; if the mean square error is less than the mean square error threshold, the iteratively adjusted purity model is used as the target purity model.

8. The method for purifying neuraminic acid based on multiple linear regression analysis according to claim 7, characterized in that: The method of using the determined coefficient to adjust the parameters of the initial purity model to obtain the iteratively adjusted purity model includes: extracting the initial coefficients preset in the initial purity model in sequence; setting a learning rate according to the determined coefficient, adjusting the initial coefficient using the learning rate to obtain the adjusted coefficient; and updating the initial coefficient using the adjusted coefficient to obtain the iteratively adjusted purity model.

9. The method for purifying neuraminic acid based on multiple linear regression analysis according to claim 8, characterized in that: The method of predicting the best purification process combination based on the target purity model includes: obtaining a purification process combination set, and extracting purification process combinations in the purification process combination set in sequence; inputting the purification process combination into the target purity model to obtain a predicted product purity value set; extracting the maximum product purity value in the predicted product purity value set, and identifying the best purification process combination corresponding to the maximum product purity value.

10. A neuraminic acid purification system based on multivariate linear regression analysis, characterized in that: The system includes: a correlation analysis and descriptive statistics module, which is used to obtain standard neuraminic acid purity data under different temperatures, pressures, solvent types and reaction times, and perform a normality test on the standard neuraminic acid purity data to obtain a maximum correlation value; a correlation analysis method is selected based on the maximum correlation value, and a correlation analysis is performed on the standard neuraminic acid purity data using the correlation analysis method to obtain a target correlation coefficient; descriptive statistics are performed on the standard neuraminic acid purity data to obtain descriptive statistical results; a verification neuraminic acid purity data preprocessing module is used to obtain verification neuraminic acid purity data, and preprocess the verification neuraminic acid purity data based on the descriptive statistics results and the target correlation coefficient to obtain target purity data; an initial purity model training module is used to establish an initial purity model using a preset multivariate linear regression analysis method, and model train the initial purity model according to the target purity data to obtain a target purity model; an optimal purification process combination prediction module is used to predict an optimal purification process combination based on a target purity model, and achieve high-purity extraction of neuraminic acid according to the optimal purification process combination.

Images