Intelligent identification method for poppy shell analogues based on electronic smell identification

Abstract

The invention provides a poppy shell analogue intelligent identification method based on electronic smell identification, and belongs to the technical field of poppy shell analogue identification, and the method comprises the steps: obtaining a to-be-identified sample, inputting the to-be-identified sample into a classification model of poppy shell analogues, and achieving the intelligent traceability analysis of the to-be-identified sample through electronic smell; the classification model of the poppy shell analogues adopts a machine learning algorithm to carry out deconstruction and comprehensive analysis construction on data obtained through HS-GC-IMS technology analysis; according to the method, experimental condition requirements are not strict, sample preparation is simple, analysis data are deconstructed and comprehensively analyzed through a machine learning algorithm, the method has the advantages of being efficient, economical and accurate in the aspect of sample traceability classification, and even nondestructive testing can be achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of identification of poppy shell analogs, and relates to an intelligent identification method for poppy shell analogs based on electronic odor identification. Background Art

[0002] In order to improve the taste of food and attract repeat customers, some illegal vendors illegally add poppy shells disguised as seasonings to food, especially in Chinese cuisine such as hot pot, marinated meat, and mutton soup. Such illegal food and drug cases occur frequently, and the accurate identification of suspected substances can provide direct evidence for case detection. Therefore, the traceability identification of poppy shell analogs is an important problem that needs to be solved urgently.

[0003] Currently, common technical methods for poppy shell analogs include spectroscopy, mass spectrometry, etc. Spectroscopic methods have shown great potential in the identification of poppy shell analogs using spectral characteristic information. Common spectroscopic methods include Fourier transform infrared spectroscopy (FTIR), Raman spectroscopy, hyperspectral imaging, etc. They are very sensitive and accurate, and can successfully identify the classification of target samples, but they require high experimental conditions, are time-consuming, and expensive. Mass spectrometry technology has a strong ability to identify the structure of unknown substances. Its combination with chromatography makes it a powerful tool for separating and identifying unknown components in poppy shell analogs, such as gas chromatography-mass spectrometry (GC-MS), liquid chromatography-mass spectrometry (LC-MS). Recently, some researchers have demonstrated that a method based on the combination of mass spectrometry and statistical tools can predict the authenticity and geographical origin of food products, with advantages such as high sensitivity, accurate qualitative analysis of unknown compounds, and precise quantitative analysis. However, this method requires extraction and concentration steps that are cumbersome, and the detection efficiency is limited.

[0004] However, in food and drug cases, the identification of poppy shell analogs places more emphasis on efficiency. Therefore, there is still a need to develop a simpler, more reliable, and more efficient method to detect the true types of suspected substances in the market. Summary of the Invention

[0005] In order to solve the technical problems existing in the above background art, the present invention provides an intelligent identification method for poppy shell analogs based on electronic odor identification, including: obtaining a sample to be identified, inputting the sample to be identified into a classification model of poppy shell analogs, and using electronic odor to realize the intelligent traceability analysis of the sample to be identified; the classification model of poppy shell analogs is constructed by deconstructing and comprehensively analyzing the data obtained by headspace gas chromatography-ion mobility spectrometry (HS-GC-IMS) technology using a machine learning algorithm.

[0006] Further, the specific steps for constructing the classification model of poppy shell analogs include: Obtaining samples, where the samples are different poppy shell analog samples; The volatile compounds in the sample are allowed to fully volatilize, and the volatile ions to be detected are separated. The volatile ions to be detected are processed to obtain a topographic map and a fingerprint map, and the unknown compounds are qualitatively analyzed to preliminarily obtain the sample classification information. A sample classification model is established, and the data is modeled, analyzed, and evaluated.

[0007] Further, the samples include, but are not limited to, tsaoko fruit, gardenia, pericarpium schizonepetae, amomum villosum, amomum kravanh, and tsaoko amomum.

[0008] Further, the sample is incubated at 60 °C and 500 rpm for 15 minutes to allow the volatile compounds to fully volatilize.

[0009] Further, the specific steps for separating the volatile ions to be detected include: The headspace vial containing the sample is placed in the instrument sample tray, and the sample is incubated at a set temperature and a set rotation speed for a set number of minutes to allow the volatile compounds to fully volatilize. The autosampler device is equipped with a sampling needle at 85 °C to automatically inject 500 μl of the top air into a headspace gas chromatography-ion mobility spectrometer (HS-GC-IMS) device, and the device model can be FlavourSpec R , G.A.S., Dortmund, Germany; The volatile analytes are separated in an MXT-5 chromatographic column, then ionized in an ion vaporization chamber, and subsequently, the volatile ions are secondarily separated in an ion migration tube according to the different drift times of the ions to obtain the volatile ions to be detected.

[0010] Further, the specific steps for preliminarily obtaining the sample classification information include: using VOCal software to process the data of the volatile compounds in the sample detected by HS-GC-IMS (including drift time, retention time, peak height, peak volume, etc.) and the data obtained by the instrument detection to obtain a topographic map and a fingerprint map, and visually presenting the distribution of the volatiles in the sample; by comparing the drift time and retention index of the compounds with the database in the instrument, the qualitative analysis of the unknown compounds can be achieved; using principal component analysis, heat map, dendrogram and other multivariate statistical analysis methods to analyze the detected data to preliminarily obtain the sample classification information.

[0011] Further, the specific steps for establishing the sample classification model include: using orthogonal partial least squares discriminant analysis to perform dimensionality reduction analysis on the data of the volatile compounds in the detected sample to preliminarily establish a sample classification model; using the random forest algorithm to perform modeling analysis on the data obtained by the instrument detection, and two-thirds of the samples are set as the calibration set to develop the classification model, and the rest are used as the validation set to evaluate the prediction ability of the model through the out-of-bag error (OOB error).

[0012] The beneficial effects of the present invention are as follows: The present invention obtains a sample to be identified, inputs the sample to be identified into a classification model of poppy shell analogs, and uses electronic odor to realize intelligent traceability analysis of the sample to be identified. The experimental conditions of this patent are not strict, and the sample preparation is simple. Through the deconstruction and comprehensive analysis of the analysis data by machine learning algorithms, it has the advantages of high efficiency, economy, and accuracy in sample traceability classification, and even non-destructive testing can be achieved.

[0013] Advantages of additional aspects of the present invention will be partially given in the following description, partially will become apparent from the following description, or will be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0015] Figure 1 It is a three-dimensional topographic map of the distribution of volatile compounds in 6 samples of the present invention; Figure 2 It is a two-dimensional spectrum of the distribution of volatile compounds in 6 samples of the present invention; Figure 3 It is a differential spectrum of the distribution of volatile compounds in 6 samples of the present invention; Figure 4 It is a fingerprint spectrum of volatile compounds in 6 samples of the present invention; Figure 5 It is a radar chart of the content distribution of major categories of volatile compounds in 6 samples of the present invention; Figure 6 It is a PCA score chart of 6 samples obtained from the differences in volatile compounds detected by HS-GC-IMS of the present invention; Figure 7 It is an OPLS-DA score chart of 6 samples obtained from the differences in volatile compounds detected by HS-GC-IMS of the present invention; Figure 8 It is a permutation test chart of the OPLS-DA model of the present invention; Figure 9 It is a clustering heat map obtained from HS-GC-IMS data of the present invention; Figure 10 It is a poppy shell analog traceability analysis chart based on the random forest algorithm of the present invention; Figure 11 It is a chart of the top 15 key volatile components of HS-GC-IMS identified by the random forest model in the training dataset of the present invention; Figure 12 Box visualization diagram of 8 key volatile components of the present invention. Detailed implementation manners

[0016] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0017] It should be noted that the following detailed description is exemplary and is intended to provide further illustration of the present invention. Unless otherwise specified, all technical and scientific terms used in this embodiment have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0018] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0019] In the present invention, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only relationship terms determined for the convenience of describing the structural relationship of each component or element of the present invention and do not specifically refer to any component or element of the present invention. It should not be construed as a limitation of the present invention.

[0020] In the present invention, terms such as "fixed connection", "connected", "connected" should be understood in a broad sense, indicating that it can be a fixed connection, an integral connection or a detachable connection; it can be directly connected or indirectly connected through an intermediate medium. For those related scientific research or technical personnel in the field, the specific meaning of the above terms in the present invention can be determined according to specific circumstances and should not be construed as a limitation of the present invention.

[0021] Embodiment 1, as Figure 1 shown, this embodiment provides an intelligent recognition method for poppy shell analogs based on electronic odor identification, including: The invention uses the emerging headspace gas chromatography-ion mobility spectrometry (HS-GC-IMS) technology to analyze volatile compounds in poppy shell analogues. This technology allows for the two-dimensional separation of volatile organic compounds directly from solid (or liquid) samples with limited or no sample preparation. After the volatile compounds in the sample are separated in the gas chromatography according to the retention time, the gas-phase ions are further secondarily separated according to their different mobilities under the action of a constant weak electric field and a countercurrent drift gas in the ion mobility tube. This tool allows for the separation and identification of isomeric compounds. This technology has low experimental conditions requirements, is economical, efficient, and has strong qualitative ability.

[0022] In addition, the invention uses machine learning algorithms to deconstruct and comprehensively analyze the data obtained by the HS-GC-IMS technology, establishes a classification model for poppy shell analogues, and realizes intelligent traceability analysis of samples using electronic odor.

[0023] Experimental steps: 1. Sample preparation: Collect different poppy shell analogue samples, such as tsaoko fruit, gardenia, pericarpium zanthoxyli, amomum villosum, amomum kravanh, tsaoko amomum, etc. Process the samples into small particles, load them into headspace sample bottles for analysis, and no pretreatment procedures such as extraction and concentration are required.

[0024] 2. Headspace gas chromatography-ion mobility spectrometry analysis: Mainly analyze the volatile compounds in the sample. Place the headspace bottle containing the sample into the instrument sample tray. First, incubate the sample at 60 °C and 500 rpm for 15 minutes to fully volatilize the volatile compounds. Then, the autosampler device equipped with a sampling needle at 85 °C automatically injects 500 μl of the top air into the device. The volatile analytes are separated in an MXT-5 chromatographic column (15 m × 0.53 mm ID, 1 μm FT, Restek, Beijing, China) at 40 °C, then ionized in the ion vaporization chamber, and subsequently, the volatile ions are secondarily separated in the ion mobility tube according to their different drift times. The carrier gas in the entire instrument separation system is 99.99% high-purity nitrogen, and the carrier gas flow rate is changed according to a specific program to achieve better separation effects.

[0025] 3. Data processing: Use VOCal software to process the data obtained by the instrument detection, and topographic distribution maps and fingerprint maps can be obtained to visually present the distribution of volatiles in the sample. By comparing the drift time and retention index of the compound with the database in the instrument, the qualitative analysis of unknown compounds can be achieved. Use multivariate statistical analysis methods such as principal component analysis (PCA), heat maps, and dendrograms to analyze the data obtained by the detection, and initially obtain sample classification information.

[0026] 4. Machine learning algorithm modeling and evaluation: Orthogonal partial least squares discriminant analysis (OPLS-DA) was used to perform dimensionality reduction analysis on the volatile compound data in the detected samples, and a sample classification model was initially established. The random forest algorithm was used to perform modeling analysis on the data obtained from the instrument detection. Two-thirds of the samples were set as the calibration set to develop the classification model, and the rest were used as the validation set to evaluate the prediction ability of the model through the out-of-bag (OOB) error. The model can accurately obtain the category information of the samples and can mark the characteristic markers in the samples. Through the above steps, by combining HS-GC-IMS analysis with machine learning algorithms, through the comprehensive analysis of volatile compounds in the samples, a classification model of poppy husk analogs was established to achieve its accurate traceability and classification.

[0027] All samples were carefully cut into small pieces for subsequent analysis. All samples were analyzed by HS-GC-IMS (FlavourSpec R , G.A.S., Dortmund, Germany), and the instrument was equipped with an automatic sampling system. Briefly described, all samples were accurately weighed 0.35 g and transferred to clean 20 ml headspace vials with magnetic screw caps (round bottom, precision screw top. The ionization of the compounds was carried out by the β-ionization source (Tritium (3H)) equipped in the HS-GC-IMS. First, the samples were incubated at 60 °C and 500 rpm for 15 minutes to allow the volatile compounds to volatilize fully. Then, the automatic sampler device equipped with a sampling needle at 85 °C automatically injected 500 μl of the top air into the device. The volatile analytes were separated in an MXT-5 chromatographic column (15 m × 0.53 mm ID, 1 μm FT, Beijing, China), then ionized in the ion vaporization chamber, and then the volatile ions were secondarily separated in the ion migration tube according to the different drift times of the ions. The carrier gas in the entire instrument separation system was 99.99% high-purity nitrogen (Jinan Deyang Special Gas Co., Ltd.), and the carrier gas flow rate control program is shown in the figure:

[0028] All poppy husk analog samples were detected in parallel 6 times to further evaluate the stability of the HS-GC-IMS method. Therefore, a total of 36 samples were detected. A control sample filled with nearly air was placed between each sample to prevent the residual volatile substances between the samples in the instrument from causing cross-contamination.

[0029] HS-GC-IMS is a two-dimensional separation method for volatile compounds. After the initial separation by gas chromatography, the ionized volatile ions are secondarily separated in the ion drift tube according to different mobilities under the action of an electric field. The principle of HS-GC-IMS can be explained by Equation 1:

[0030] K represents the mobility value, which is used to determine the reduced ion mobility characterizing the volatile organic components. e represents the unit charge, and N represents the number density of the drift gas. Then two parameters m and M are calculated, representing the ionic mass and molecular weight of the analyte, respectively. Where k represents the Boltzmann constant, r represents the minimum value in the potential curve, and “Ω” is the one-dimensional collision cross section.

[0031] The qualitative analysis of volatile compounds is carried out by comparing their retention indices and drift times with those of compounds in the HS-GC-IMS database. The data collected by HS-GC-IMS are processed by VOCal software (G.A.S., Germany) containing several plugins, including reporter, galerie, and Dynamic PCA. The Reporter plugin is used to explore 2D and 3D topographic maps and differential spectra. The Gallery plugin is used to generate fingerprint spectra.

[0032] 2.3 Data Processing The raw data obtained by HS-GC-IMS are further processed by multivariate statistical analysis and machine learning algorithms. Multivariate statistical analyses such as principal component analysis (PCA), volcano plot, orthogonal partial least squares discriminant analysis (OPLS-DA), hierarchical clustering analysis (HCA), and clustering heatmap are obtained using SIMCA software (version 14.1, Umetrics, Umea, Sweden). The clustering heatmap is obtained by TBtools software. Random forest (RF), receiver operating characteristic curve (ROC), and volcano plot are obtained by MetaboAnalyst 6.0 software (http: / / www.metaboanalyst.ca). Before analysis using machine learning algorithms, all data are standardized by the pareto-scaling method.

[0033] 3. Results and Discussion, 3.1 Volatile Characteristics of Poppy Shell Analogs, The volatiles in the samples are vividly shown in Figure 1 which intuitively shows the types and content characteristics of volatiles in different samples. Figure 1It is a 3D topographic map. The x-axis represents the migration time of ions in the drift tube, the y-axis represents the retention time in the gas chromatography column, and the z-axis represents the signal intensity. Higher peaks represent higher compound concentrations. The red vertical line on the left side of the figure is the reaction ion peak, which represents the total ion concentration of water ionized in the tritium ion source. Protonated water can further combine with volatile molecules in the sample and then ionize the volatile molecules.

[0034] To obtain a more intuitive image, the information of the volatile content represented by the peak height on the z-axis in the 3D map is represented by the brightness of the color, thus obtaining a 2D image, which is shown in Figure 2 . Blue represents the background signal. Each bright spot represents a volatile compound. The redder the color, the higher the concentration of the volatile, and the whiter the color, the lower the concentration of the volatile. The signals of the volatiles in the 2D topographic map are mainly concentrated in the range of retention index 100 - 1200, indicating that there are a large number of small polar volatile molecules in the poppy shell analog sample. To visually compare the differences in volatiles in different poppy shell analog samples, the differential spectrum is shown in Figure 3 . This figure is obtained by comparing the contents of the same volatiles in different samples. Based on this, it can be deduced that the white background indicates similar volatile contents, while red indicates that the concentration of this volatile compound is higher compared to the reference sample. Correspondingly, blue indicates that the concentration of the volatile compound is lower compared to the reference sample. In the range of retention index 100 - 800, compared with the reference sample XS, the volatile contents of SR, CG, YKZ, and ZZ increase, as shown by the large number of red dots in the purple dashed box in Figure 3 . In addition, the spectra of SR and CG are relatively similar. Compared with other samples, they have more volatile compounds in the range of 900 - 1600, as shown in the green dashed box in Figure 3 . This similarity is related to the fact that SR and CG come from the same genus, and it also shows that using HS - GC - IMS to detect the components of volatile compounds in samples is beneficial for distinguishing the sources of samples.

[0035] 3.2 Qualitative analysis of volatile compounds and fingerprint analysis of poppy shell analogs. HS - GC - IMS is a two - dimensional separation technique. Volatile compounds are first separated once in the gas chromatography, and then the volatile ions ionized in the ionization chamber are secondarily separated in the drift tube under the action of an electric field. Therefore, as shown in Figure 1Among them, each volatile has three-dimensional information, namely retention time, drift time, and signal intensity. Subsequently, the volatile substances in the sample were identified by comparing the retention index of the compound with the drift time and using the GC × IMS database search. By analyzing the peak volume of the volatiles in the spectrogram, their quantitative information can be obtained. A total of 255 volatile compounds were detected in 6 samples of poppy shell analogs from different sources. It should be noted that when the concentration of the compound is high enough, dimers or even trimers will be formed, and the formation of polymers is also related to the proton affinity of the volatile substances.

[0036] To more clearly visualize the differences in volatiles between different samples, fingerprint maps were created to solve this difficulty, as Figure 4 shown. The volatile compounds detected in 6 samples of poppy shell analogs are shown in boxes, and the sample types are distinguished by rows, while the compound types are distinguished by columns. The brightness of the square represents the concentration of the compound, and the redder the color, the higher the content of this compound in the sample. The characteristic fingerprint map can distinguish poppy shell analogs based on the types and contents of the compounds. In the fingerprint map, each sample is arranged and clustered according to its unique volatile compounds, and this ladder diagram effectively highlights the differences in volatiles between samples.

[0037] Generally speaking, the HS-GC-IMS fingerprint map can intuitively display the similarities and differences of volatile compounds between different samples at the same time. It is an effective method for comprehensively investigating the components of samples and can distinguish 6 poppy shell analogs through the fingerprint map.

[0038] 3.3 Content analysis of volatiles in poppy shell analogs. It can be observed from the left side of the fingerprint map that some volatiles exist in all 6 samples of poppy shell analogs, but their contents are different. At this time, the analysis based on the compound concentration is an important basis for the accurate classification of samples. Therefore, in this study, semi-quantitative analysis was carried out on various compounds based on the peak volume of the volatiles in HS-GC-IMS. The radar map intuitively and clearly shows the content distribution of various volatiles in 6 samples, as Figure 5 . The 10 directions in the figure represent different categories of compounds, the lines of different colors represent the sample types, and the points of different colors on the coordinate of each direction represent the content of this type of compound in the corresponding sample.

[0039] HS-GC-IMS can clearly identify the differences in the types and contents of volatiles in the sample and accurately identify the types of poppy shell analogs. More importantly, by comprehensively analyzing the types and contents of volatiles in the sample, the medicinal value of the sample can be further evaluated and explored, indicating that the HS-GC-IMS technology has great potential application value in the development of medicinal value.

[0040] 3.4 Multivariate statistical analysis of poppy shell analogues from different sources. The volatile results data obtained by HS-GC-IMS technology is very complex. The spectra not only include information such as retention index, retention time, and peak volume, but also include a large number of characteristic signals and non-characteristic signals, resulting in difficulties in the precise classification of samples. To deeply explore the correlation between data characteristics and sample classification, PCA was used to establish a classification model. PCA is an unsupervised classification model that transforms the original complex HS-GC-IMS data into a set of independent linear representations in each dimension through linear transformation, achieving dimensionality reduction of high-dimensional data. Figure 6 The PCA score plot of 6 poppy shell analogues is given, which can aggregate and classify the 6 samples in an unsupervised mode. However, in the unsupervised PCA score plot, the scores of SR and CG in both PC1 and PC2 dimensions are relatively close, resulting in their inability to be clearly distinguished, indicating that there are certain limitations in the unsupervised PCA classification model when identifying species with relatively close biological relationships. Therefore, a more precise and intelligent method is needed for further analysis.

[0041] 3.5 OPLS-DA machine learning model. OPLS-DA is a supervised machine learning classification method that removes the data variation in independent variable X that is irrelevant to classification variable Y through orthogonal signal correction (OSC) technology. It makes the classification information mainly concentrated in one principal component, and then searches for the direction of the orthogonal correction axis of this principal component. It weakens the within-group differences and maximally highlights the between-group differences, so that the separation effect of between-group samples is better. The OPLS-DA score plot of poppy shell analogues is as Figure 7 shown. It can be seen from this figure that the 6 samples are distributed in different quadrants and are clearly separated. It should be noted that both SR and CG samples are distributed in the second quadrant. Although the scores of their abscissas are similar, the scores of their ordinates are quite different, and the two groups of samples can be clearly clustered and separated. This shows that compared with the unsupervised PCA model, the supervised OPLA-DA machine learning model has a better classification effect.

[0042] The classification and prediction ability of the OPLS-DA model is evaluated by R 2 Y and Q 2 (cum). When R 2 Y and Q 2 exceeds 0.5, it indicates that the fitting result of the model is acceptable, and the closer their values are to 1, the better the fitting degree of the model, and the more accurately the samples in the training set can be classified into their original categories. The R 2 Y and Q 2They are 0.986 and 0.98 respectively, indicating that the model has excellent predictive ability for the classification of poppy shell analog samples. The disadvantage of the supervised classification model is that over-fitting may occur. Therefore, we used 200 permutation tests to evaluate the rationality and reliability of the model. As Figure 8 shown, the x-axis represents the permutation retention degree, and its value represents the similarity with the original model. The y-axis represents R 2 / Q 2 , and the dotted line is the regression line. After 200 cross-validations, the regression line of R 2 intersects the Y-axis at (0.0, 0.0511), which is significantly less than the model variable interpretability and < 0.1, and the intercept of the Q 2 regression line < 0 (Q 2 = (0.0, -0.442)), and all the simulated values on the left are less than the true value on the far right. The above evidence shows that the OPLS-DA model does not have over-fitting, is reliable, and has good predictive ability for unknown samples. The above shows that the OPLS-DA model can be used to discover the signature volatile compounds in poppy shell analogs and trace and identify different samples.

[0043] To more clearly show the relationship between the types and abundance differences of volatile compounds and sample classification, a clustering heatmap was used for further interpretation, as Figure 9 shown. Each latitude represents a sample, and each longitude represents a volatile compound. The heatmap clearly shows the differences in the types and contents of volatile compounds among different samples. The change in its color from blue to red represents the increase in the concentration of volatile compounds. All poppy shell analog samples were clearly and accurately clustered into 6 categories, and the information of characteristic volatile compounds was consistent with the fingerprint spectrum ( Figure 4 ). The above conclusions further illustrate that the types and contents of volatile compounds play a crucial role in the traceability analysis of poppy shell analogs.

[0044] 3.6 Traceability analysis of poppy shell analogs based on the random forest algorithm. As described in the fingerprint spectrum, the original data obtained by HS-GC-IMS is very complex, showing many redundant features, which is not conducive to the accurate classification of samples. Therefore, to further improve the discrimination efficiency of samples, a precise machine learning classification model was established in this study. This model uses the random forest algorithm combined with HS-GC-IMS to characterize and explore the volatile compounds in poppy shell analog samples. It is a supervised method based on an ensemble strategy. In fact, the random forest algorithm is an ensemble method that uses a series of decision trees for modeling. By integrating the prediction results of multiple decision trees, it can reduce the over-fitting of the model, thereby improving the accuracy and stability of the model.

[0045] Two-thirds of the samples were set as the calibration set to develop the classification model, and the rest were used as the validation set to evaluate the predictive ability of the model through the out-of-bag (OOB) error. Generally, training with 500 trees is sufficient to stabilize the OOB error. The random forest classification model is as follows Figure 10 shown. The abscissa represents the number of decision trees, and the ordinate represents the OOB error. After approximately 30 decision trees, the OOB error approaches stability, and its value reaches 0 according to the majority voting principle. In addition, according to the mean decrease in accuracy, a total of 15 compounds were selected. They would greatly increase the OOB error and were thus evaluated as important differential volatiles for distinguishing 6 different poppy shell analogs, as shown in Figure 11 shown.

[0046] Figure 12 a-h show the normalized relative contents of the top 8 significant features in 6 samples. This is the result mainly generated by the random forest algorithm based on the mean decrease in accuracy.

[0047] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent recognition method for poppy shell analogs based on electronic odor identification, characterized in that, Including: Obtain the sample to be identified, input the sample to be identified into the classification model of papaver somniferum shell analogues, and use electronic odor to realize the intelligent traceability analysis of the sample to be identified; The classification model of papaver somniferum shell analogues is constructed by using machine learning algorithms to deconstruct and comprehensively analyze the data obtained by HS-GC-IMS technology.

2. The intelligent recognition method of poppy shell analog based on electronic odor identification according to claim 1, characterized in that The specific steps for constructing the classification model of papaver somniferum shell analogues include: Obtain samples, and the samples are different papaver somniferum shell analogue samples; Fully volatilize the volatile compounds in the sample, and separate to obtain the volatile ions to be detected; Process the volatile ions to be detected to obtain a topographic map and a fingerprint map, qualitatively analyze the unknown compounds, and initially obtain the sample classification information; Establish a sample classification model, and perform modeling analysis and evaluation on the data.

3. The intelligent recognition method of poppy shell analogues based on electronic odor identification according to claim 2, characterized in that The samples include but are not limited to tsaoko fruit, gardenia, pericarpium citri reticulatae, fructus amomi, semen amomi, and tsaoko amomum fruit.

4. The intelligent recognition method of poppy shell analog based on electronic odor identification according to claim 2, characterized in that, The sample is incubated at 60 °C and 500 rpm for 15 minutes to fully volatilize the volatile compounds.

5. The intelligent recognition method of poppy shell analogues based on electronic odor identification according to claim 2, characterized in that, The specific steps for separating to obtain the volatile ions to be detected include: Place the headspace vial containing the sample into the instrument sample tray, and incubate the sample at the set temperature and set rotation speed for the set number of minutes to fully volatilize the volatile compounds; The autosampler device is equipped with a sampling needle at 85 °C to automatically inject 500 μl of the top air into the device; The volatile analytes are separated in the MXT-5 chromatographic column, then ionized in the ion vaporization chamber, and subsequently, the volatile ions are secondarily separated according to the different drift times of the ions in the ion migration tube to obtain the volatile ions to be detected.

6. The intelligent recognition method of poppy shell analogues based on electronic odor identification according to claim 2, characterized in that The specific steps for initially obtaining the sample classification information include: using VOCal software to process the data obtained by the instrument detection to obtain a topographic map and a fingerprint map, and visually presenting the distribution of volatile substances in the sample; by comparing the drift time and retention index of the compound with the database in the instrument, the qualitative analysis of the unknown compound can be realized; using multivariate statistical analysis methods such as principal component analysis, heat map, and phylogenetic tree diagram to analyze the data obtained by the detection, and initially obtaining the sample classification information.

7. The intelligent recognition method of poppy shell analog based on electronic odor identification according to claim 2, characterized in that, The specific steps for establishing the sample classification model include: using orthogonal partial least squares discriminant analysis to perform dimensionality reduction analysis on the data of volatile compounds in the detected samples, and initially establishing the sample classification model; using the random forest algorithm to perform modeling analysis on the data obtained by the instrument detection, two-thirds of the samples are set as the calibration set to develop the classification model, and the rest are used as the validation set to evaluate the prediction ability of the model through the out-of-bag error (OOB error).

Images