Maize freshness classification method based on machine learning algorithm

Abstract

The invention discloses a corn freshness classification method based on a machine learning algorithm. The method comprises the following steps: acquiring common volatile component data of new corn and old corn; using a machine learning algorithm to train the volatile component data to obtain a corn new aging degree classification model; and analyzing volatile component data of a to-be-detected corn kernel sample according to the corn freshness classification model to obtain a classification result. According to the corn freshness classification method provided by the invention, the freshness of unknown samples can be predicted, the prediction accuracy is relatively high, a reference thought is provided for developing a corn freshness identification method, and the corn freshness classification method has important significance in guaranteeing national grain safety and saving grain loss.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of analysis and detection, and specifically to a corn freshness classification method based on a machine learning algorithm. Background Art

[0002] Corn is one of the world's major food crops and an important source of food, feed, fuel and industrial raw materials. China is the world's second largest corn producer and major consumer. As one of China's three major staple crops, corn is mainly grown in Northeast China, North China and parts of Southwest China. Corn provides important nutrients for the human body, such as starch, protein and fat, and is rich in polyphenols, carotenoids, vitamins and dietary fiber. In order to meet the growing demand for food, ensure national food security and promote global sustainable development, it is necessary to implement storage measures for corn. During storage, corn will be affected by environmental factors, causing the physical and chemical properties to gradually change, which in turn affects the edible quality. Therefore, it is of great significance to study rapid, accurate and non-destructive storage quality identification methods.

[0003] Changes in volatile components are important characteristics that indirectly reflect changes in storage quality and nutritional value of corn. However, olfactory perception alone cannot fully identify corn quality. Therefore, it is necessary to develop more accurate and reliable methods to analyze volatile components in stored corn to monitor its quality changes. Solid phase microextraction can efficiently extract volatile compounds from samples in a simple, rapid and highly sensitive manner, and when coupled with gas chromatography-mass spectrometry, it helps to accurately identify flavor components in cereals. Non-targeted flavoromics methods using solid phase microextraction-gas chromatography-mass spectrometry have been used to analyze volatile substances in various cereals, including wheat, sweet corn, rice, black sesame and soybeans. Although non-targeted methods have a wide coverage, they have the disadvantages of poor repeatability and low detection limit. In contrast, broadly targeted methods have higher sensitivity, specificity and superior quantitative capabilities. Summary of the invention

[0004] To this end, an embodiment of the present invention provides a method for classifying corn freshness based on a machine learning algorithm. The present invention uses the multiple reaction monitoring (MRM) mode of solid phase microextraction-gas chromatography-mass spectrometry (SPME-GC-MS / MS) to establish a targeted detection method for volatile components. The method is used to detect corn harvested from the seven major producing provinces in northern China from 2019 to 2023, and three machine learning algorithms, support vector machine, multi-layer perception and gradient boosting, are used to construct a corn freshness classification model. The results show that SPME-GC-MS / MS technology is very suitable for non-destructive detection of volatile components in stored corn, and the established corn freshness classification model has high accuracy, precision, recall and prediction accuracy.

[0005] In order to achieve the above purpose, the embodiment of the present invention provides the following technical solutions:

[0006] A method for classifying corn freshness based on a machine learning algorithm, the method comprising:

[0007] Obtain volatile component data common to new and old corn;

[0008] Using a machine learning algorithm to train the volatile component data to obtain a corn freshness classification model;

[0009] The volatile component data of the corn kernel sample to be tested is analyzed according to the corn freshness classification model to obtain a classification result.

[0010] Furthermore, the new corn refers to corn harvested in the current year, and the old corn refers to corn that has been stored for more than one year.

[0011] Further, the common volatile components are: 2-butanone, 2-ethylfuran, 2,3-butanedione, dimethyl disulfide, hexanal, 4-methyl-3-penten-2-one, n-butanol, methyl hexanoate, isopentanol, 2-methylpyridine, trans-2-hexenal, 2-pentylfuran, 2-methylpyrazine, 2-ethylpyridine, sec-octanone, n-octanal, 1-octen-3-one, 2,5-dimethylpyrazine, trans-2-heptenal, 6-methyl-5-hepten-2-one, n-Hexanol, dimethyl trisulfide, methyl octanoate, nonanal, 2,3,5-trimethylpyrazine, trans-3-octen-2-one, trans-2-octenal, cis-linalool oxide, 1-octen-3-ol, n-heptanol, acetic acid, furfural, trans-linalool oxide, 2-ethylhexanol, decanal, 2-acetylfuran, benzaldehyde, ethyl nonanoate, trans-2-nonanal, propionic acid, octanol, isobutyric acid, 2-pentylpyridine, 5-methylfurfural, trans, cis-2,6-nonadienal, methyl 1-Nonyl Ketone, Undecyl Alcohol, Trans-2-Octene-1-ol, Benzonitrile, Butyric Acid, Trans-2-Decenyl Alcohol, Phenylacetaldehyde, Acetophenone, 1-Nonanol, Furfuryl Alcohol, Isopentyl Acid, Trans-2,4-Nonadienal, γ-Caprolactone, Dodecanal, N-Valeric Acid, Trans-2-Undecenal, N-Decanol, 4'-Methylacetophenone, Methyl Salicylate, 4-Isopropylbenzaldehyde, γ-Heptanolactone, Trans-2,4-Decadienal, Hexanoic Acid, Geranyl Acetone, Guaiacol, Butyl Benzoate, Benzyl alcohol, phenylethyl alcohol, γ-octalactone, β-ionone, dodecanol, benzothiazole, 2-acetylpyrrole, o-cresol, phenol, 4-ethyl-2-methoxyphenol, p-methoxybenzaldehyde, γ-nonalactone, pantolactone, trans-cinnamaldehyde, octanoic acid, p-cresol, hexadecanal, γ-decanoic acid, nonanoic acid, 4-ethylphenol, p-vinylguaiacol, o-aminoacetophenone, massoia lactone, n-decanoic acid, indole, 3-methylindole, vanillin, vanillyl acetone.

[0012] Furthermore, the method for acquiring the volatile component data includes:

[0013] Solid phase microextraction-gas chromatography-triple quadrupole mass spectrometry was used to detect corn kernel samples and obtain volatile component data.

[0014] Furthermore, the parameters of the solid phase microextraction-gas chromatography-triple quadrupole mass spectrometry technology are as follows:

[0015] The extraction head is Smart SPMEArrow;

[0016] The pre-aging condition of the extraction head was 250°C for 5 min;

[0017] Extraction conditions: temperature 70°C, time 60 min;

[0018] The desorption condition was 250 °C for 2 min;

[0019] The chromatographic column is SH-PolarWax;

[0020] The injection port temperature was 250 °C;

[0021] Column flow rate: 1.00 mL / min;

[0022] The injection mode was splitless;

[0023] The temperature program was as follows: the initial column temperature was maintained at 40 °C for 5 min, then gradually increased to 250 °C at a rate of 3 °C / min and maintained for 15 min;

[0024] The ionization source is an electron bombardment ionization source with an electron energy of 70 eV;

[0025] The ion source temperature was 200 °C;

[0026] The interface temperature is 250°C;

[0027] Solvent delay 5min;

[0028] The data acquisition method is simultaneous acquisition of multiple reaction monitoring and full scan;

[0029] The mass scan range was 33–550 m / z;

[0030] The mass spectral library is the Shimadzu aroma database.

[0031] Furthermore, the machine learning algorithm includes support vector machine, multi-layer perception and gradient boosting.

[0032] The embodiments of the present invention have the following advantages:

[0033] 1. The present invention adopts the MRM mode of SPME-GC-MS / MS to establish a targeted detection method for volatile components of corn, analyzes the volatile components of corn samples in different storage years, and obtains the volatile components common to new and old corn.

[0034] 2. The corn freshness classification method based on the machine learning algorithm of the present invention can realize the prediction of the freshness of unknown samples with high prediction accuracy, which provides a reference idea for the development of corn freshness identification methods and is of great significance to ensuring national food security and saving grain and reducing losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0036] Figure 1 A classification diagram of volatile components in corn provided by the present invention;

[0037] Figure 2 A graph of the relative contents of different types of volatile components in corn harvested from 2019 to 2023 provided by the present invention;

[0038] Figure 3 The evaluation result of the corn freshness classification model constructed based on the support vector machine provided by the present invention;

[0039] Figure 4 The evaluation result of the corn freshness classification model based on multi-layer perception provided by the present invention;

[0040] Figure 5 This is the evaluation result of the corn freshness classification model constructed based on gradient boosting provided by the present invention. DETAILED DESCRIPTION

[0041] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Example 1

[0043] 1. Materials and Methods

[0044] 1.1 Sample collection

[0045] A total of 285 corn kernel samples collected from the field with harvest years from 2019 to 2023 were collected. The origins included seven major corn producing areas in northern China. The sample geographical information, sampling year and sample quantity are shown in Table 1.

[0046] Table 1. Overview of corn kernel samples

[0047]

[0048] 1.2 Experimental instruments

[0049] Shimadzu GCMS-TQ 8040 gas chromatograph-triple quadrupole mass spectrometer, equipped with AOC-6000plus multi-function autosampler.

[0050] 1.3 Sample analysis

[0051] Weigh 6.0 g corn kernels into a 20 mL headspace sample bottle, add 2-methyl-3-heptanone solution (0.1 μg / mL, 10 μL), seal with a magnetic lid and place on a sample rack;

[0052] The sample bottle was thermally balanced at 70°C for 20 min, and the volatile components were extracted using a Smart SPMEArrow extraction head (DVB / CAR WR / PDMS coating, outer diameter 1.10 mm, stationary phase thickness 120 μm, extraction head pre-aged at 250°C for 5 min) at 70°C for 60 min, followed by desorption at 250°C for 2 min;

[0053] An SH-PolarWax chromatographic column (60m×0.25mm×0.25μm) was used to separate the volatile components. The injection port temperature was 250℃, the column temperature program was 40℃ for 5min, then increased to 250℃ at 3℃ / min and maintained for 15min, the carrier gas control mode was constant linear velocity mode, and the column flow rate was 1.00mL / min; the injection mode was splitless injection; the ion source temperature was 200℃, the interface temperature was 250℃, and the solvent delay was 5min; the acquisition mode was multiple reaction monitoring mode (MRM) and full scan simultaneous acquisition, and the mass scan range was 33-550m / z.

[0054] 1.4 Statistical analysis

[0055] Qualitative analysis: The Shimadzu Aroma Smart Database was used to analyze the volatile components in corn samples. The Aroma_TQ_SH-PolarWax_AART method in the database package was used to determine the n-alkane samples and perform peak identification to calibrate the retention time of the target components. The MRM method file was established using the n-alkane data and aroma database, and the volatile components in corn were screened and identified in combination with the mass spectra and characteristic ion information of the compounds in the NIST 17.0 mass spectral library.

[0056] Quantitative analysis: 2-methyl-3-heptanone was used as the internal standard substance, and the peak area ratio normalization method was used to calculate the relative content of each volatile flavor substance detected in corn. The formula is as follows:

[0057]

[0058] Where: C i is the content of the detected substance μg / kg, C0 is the concentration of the internal standard μg / mL, V0 is the volume of the internal standard μL, A i is the chromatographic peak area of ​​the detected substance, A0 is the chromatographic peak area of ​​the internal standard, and m is the sample mass in g.

[0059] GCMSsolution software was used to extract the data, and the retention time and peak area of ​​each volatile flavor substance were obtained by peak integration. The data were exported to Microsoft Excel 2019 software for data processing. Origin 2019 software was used for statistical analysis and mapping of the data.

[0060] 1.5 Machine Learning

[0061] Dataset: A database was constructed using 285 corn samples as modeling objects, and features were selected from the 99 volatile components identified by qualitative analysis for classifier training. 29 samples were randomly selected from the 285 corn samples as the validation set, and the remaining samples were divided into training and test sets at a ratio of 8:2, i.e., 205 samples in the training set and 51 samples in the test set.

[0062] Data preprocessing: The data were normalized and standard scaled using a self-developed Python program on an Anaconda Jupyter notebook (version 6.5.4), and then the feature dimensionality was reduced using the PLS regression algorithm.

[0063] Classifier training: Based on 205 sample data, three algorithms, support vector machine (SVM), gradient boosting (GBC) and multi-layer perception (MLP), were used to train the classifier to obtain the corn freshness classification model. The performance of the constructed corn freshness classification model was evaluated using the test set and validation set to ensure the accuracy and reliability of the model. The stability of the corn freshness classification model was evaluated using 10-fold cross validation. The training set was divided into 10 groups, 9 groups of data were used to train the classifier, and the remaining 1 data set was used to evaluate the corn freshness classification model.

[0064] 2. Results and Analysis

[0065] 2.1 Analysis of volatile components of corn

[0066] A total of 99 common volatile components were detected in 285 corn samples, which can be divided into 8 categories, including 24 aldehydes, 16 alcohols, 14 ketones, 13 heterocyclic compounds, 12 esters, 10 acids, 7 phenols and 4 other compounds. It can be seen that aldehydes are the main volatile components in corn, accounting for 24.2% of the total volatile components. The threshold of aldehydes is low, and they are the main products of fat oxidation, which have an important impact on the overall flavor of corn; the number of alcohols accounts for 16.2% of the total flavor substances, and they are the secondary products of unsaturated fatty acid oxidation; ketones mainly come from the oxidation or thermal degradation of unsaturated fatty acids, as well as the degradation of amino acids or microbial oxidation. The threshold of ketones is high and has little effect on the overall flavor of corn. In addition, the relative content of volatile components in corn changes with the increase of storage time. The content of acids is relatively high, followed by aldehydes and alcohols, and the content of aldehydes and alcohols is relatively high in freshly harvested corn. The common volatile components are: 2-butanone, 2-ethylfuran, 2,3-butanedione, dimethyl disulfide, hexanal, 4-methyl-3-penten-2-one, n-butanol, methyl hexanoate, isopentanol, 2-methylpyridine, trans-2-hexenal, 2-pentylfuran, 2-methylpyrazine, 2-ethylpyridine, sec-octanone, n-octanal, 1-octen-3-one, 2,5-dimethylpyrazine, trans-2-heptenal, 6-methyl-5-hepten-2-one, n-hexanol, Dimethyl trisulfide, methyl octanoate, nonanal, 2,3,5-trimethylpyrazine, trans-3-octen-2-one, trans-2-octenal, cis-linalool oxide, 1-octen-3-ol, n-heptanol, acetic acid, furfural, trans-linalool oxide, 2-ethylhexanol, decanal, 2-acetylfuran, benzaldehyde, ethyl nonanoate, trans-2-nonanal, propionic acid, octanol, isobutyric acid, 2-pentylpyridine, 5-methylfuranal, trans, cis-2,6-nonadienal, methylnonyl Ketone, undecyl aldehyde, trans-2-octen-1-ol, benzonitrile, butyric acid, trans-2-decenal, phenylacetaldehyde, acetophenone, 1-nonanol, furfuryl alcohol, isovaleric acid, trans, trans-2,4-nonadienal, γ-caprolactone, dodecanal, n-valeric acid, trans-2-undecenal, n-decanol, 4'-methylacetophenone, methyl salicylate, 4-isopropylbenzaldehyde, γ-heptanolactone, trans-2,4-decadienal, hexanoic acid, geranylacetone, guaiacol, butyl benzoate, benzyl Alcohol, phenylethyl alcohol, γ-octalactone, β-ionone, dodecanol, benzothiazole, 2-acetylpyrrole, o-cresol, phenol, 4-ethyl-2-methoxyphenol, p-anisaldehyde, γ-nonalactone, pantolactone, trans-cinnamaldehyde, octanoic acid, p-cresol, hexadecanal, γ-decanoic acid, nonanoic acid, 4-ethylphenol, p-vinylguaiacol, o-aminoacetophenone, massoia lactone, n-decanoic acid, indole, 3-methylindole, vanillin, vanillone.

[0067] 2.2 Corn freshness classification based on machine learning

[0068] SPME-GC-MS / MS was used to analyze the peak area of ​​volatile components in corn samples with different storage times. The corn harvested in 2023 was regarded as new corn, and the corn harvested in 2019-2022 was regarded as old corn. After grouping, the sample volatile component data was imported into a self-compiled Python program. Table 2 lists the test set accuracy, precision rate, recall rate and prediction accuracy of the validation set of the trained classifier. It can be seen that the test set accuracy, precision rate, recall rate and cross-validation rate of the three algorithms of support vector machine, multi-layer perception and gradient boosting obtained by the present invention are all 100%, and the prediction accuracy of the validation set reaches more than 95%.

[0069] Table 2. Classification performance of trained classifiers

[0070]

[0071] 3. Conclusion

[0072] The volatile components of 285 corn samples from different storage years were analyzed using SPME-GC-MS / MS and MRM broad targeted detection analysis methods. The optimal SPME conditions for volatile flavor compounds in corn are: sample dosage 6.0g, extraction temperature 70°C, and extraction time 60 minutes. Based on the above pretreatment conditions, a total of 99 common flavor compounds were detected, including 8 categories of compounds such as aldehydes, alcohols, esters and phenols. Based on the above 99 common flavor compounds, the present invention established a machine learning method that can accurately distinguish between new and old corn. Compared with human senses, the volatile omics method is more accurate and reliable in analyzing the unique flavor characteristics of corn, and provides scientific technical means and theoretical support for monitoring the storage quality of corn.

[0073] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A corn freshness classification method based on machine learning algorithm, characterized in that: The method comprises: Obtain volatile component data common to new and old corn; Using a machine learning algorithm to train the volatile component data to obtain a corn freshness classification model; The volatile component data of the corn kernel sample to be tested is analyzed according to the corn freshness classification model to obtain a classification result.

2. The corn freshness classification method based on machine learning algorithm according to claim 1, characterized in that: The new corn refers to corn harvested in the current year, and the old corn refers to corn that has been stored for more than one year.

3. The corn freshness classification method based on machine learning algorithm according to claim 1, characterized in that: The common volatile components are: 2-butanone, 2-ethylfuran, 2,3-butanedione, dimethyl disulfide, hexanal, 4-methyl-3-penten-2-one, n-butanol, methyl hexanoate, isopentanol, 2-methylpyridine, trans-2-hexenal, 2-pentylfuran, 2-methylpyrazine, 2-ethylpyridine, sec-octanone, n-octanal, 1-octen-3-one, 2,5-dimethylpyrazine, trans-2-heptenal, 6-methyl-5-hepten-2-one, n-hexanol, Dimethyl trisulfide, methyl octanoate, nonanal, 2,3,5-trimethylpyrazine, trans-3-octen-2-one, trans-2-octenal, cis-linalool oxide, 1-octen-3-ol, n-heptanol, acetic acid, furfural, trans-linalool oxide, 2-ethylhexanol, decanal, 2-acetylfuran, benzaldehyde, ethyl nonanoate, trans-2-nonanal, propionic acid, octanol, isobutyric acid, 2-pentylpyridine, 5-methylfuranal, trans, cis-2,6-nonadienal, methylnonyl Ketone, undecyl aldehyde, trans-2-octen-1-ol, benzonitrile, butyric acid, trans-2-decenal, phenylacetaldehyde, acetophenone, 1-nonanol, furfuryl alcohol, isovaleric acid, trans, trans-2,4-nonadienal, γ-caprolactone, dodecanal, n-valeric acid, trans-2-undecenal, n-decanol, 4'-methylacetophenone, methyl salicylate, 4-isopropylbenzaldehyde, γ-heptanolactone, trans-2,4-decadienal, hexanoic acid, geranylacetone, guaiacol, butyl benzoate, benzyl Alcohol, phenylethyl alcohol, γ-octalactone, β-ionone, dodecanol, benzothiazole, 2-acetylpyrrole, o-cresol, phenol, 4-ethyl-2-methoxyphenol, p-anisaldehyde, γ-nonalactone, pantolactone, trans-cinnamaldehyde, octanoic acid, p-cresol, hexadecanal, γ-decanoic acid, nonanoic acid, 4-ethylphenol, p-vinylguaiacol, o-aminoacetophenone, massoia lactone, n-decanoic acid, indole, 3-methylindole, vanillin, vanillone.

4. The corn freshness classification method based on machine learning algorithm according to claim 1, characterized in that: The method for obtaining the volatile component data comprises: Solid phase microextraction-gas chromatography-triple quadrupole mass spectrometry was used to detect corn kernel samples and obtain volatile component data.

5. The corn freshness classification method based on machine learning algorithm according to claim 4 is characterized in that: The parameters of the solid phase microextraction-gas chromatography-triple quadrupole mass spectrometry technology are as follows: The extraction head is Smart SPMEArrow; The pre-aging condition of the extraction head was 250°C for 5 min; Extraction conditions: temperature 70°C, time 60 min; The desorption condition was 250 °C for 2 min; The chromatographic column is SH-PolarWax; The injection port temperature was 250 °C; Column flow rate: 1.00 mL / min; The injection mode was splitless; The temperature program was as follows: the initial column temperature was maintained at 40 °C for 5 min, then gradually increased to 250 °C at a rate of 3 °C / min and maintained for 15 min; The ionization source is an electron bombardment ionization source with an electron energy of 70 eV; The ion source temperature was 200 °C; The interface temperature is 250°C; Solvent delay 5min; The data acquisition method is simultaneous acquisition of multiple reaction monitoring and full scan; The mass scan range was 33–550 m / z; The mass spectral library is the Shimadzu aroma database.

6. The corn freshness classification method based on machine learning algorithm according to claim 1, characterized in that: The machine learning algorithms include support vector machine, multi-layer perception and gradient boosting.

Images