Tan mutton production place identification method based on back propagation neural network

Abstract

The invention discloses a Tan mutton producing area identification method based on a back-propagation neural network, the back-propagation neural network is optimized through a particle swarm optimization algorithm, and the method comprises the following steps: extracting lipids in Tan mutton samples from different producing areas; an LC-MS / MS system is adopted to analyze the components and contents of lipids in the Tan mutton samples from different producing areas; screening the components and contents of lipids in all the samples by adopting a random forest model to obtain characteristic lipid data of the beach mutton samples from different producing areas; training a pre-constructed back-propagation neural network optimized based on a particle swarm algorithm by adopting the characteristic lipid data of the Tan mutton samples from different producing areas; and identifying the producing area of the Tan mutton with an unknown producing area by adopting the trained back propagation neural network optimized based on the particle swarm algorithm. According to the method, the characteristic lipids of the Tan mutton in different producing areas are screened out through the random forest algorithm, and accurate identification of the producing areas of the Tan mutton in different areas is realized through the back-propagation neural network optimized by the particle swarm optimization algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of food origin identification, and more specifically, to a method for identifying the origin of Tan mutton based on a back propagation neural network. Background Art

[0002] At present, the main detection technologies used for the origin traceability of Tan mutton include stable isotope method, mineral element method, nuclear magnetic resonance spectroscopy, infrared spectroscopy fingerprint method, organic component fingerprint analysis method and other methods. However, these traditional Tan mutton origin traceability detection technologies have the disadvantages of complex pre-processing or large sampling volume. It is urgent to develop a fast and reliable Tan mutton origin traceability method with small sampling volume to maintain market fairness and protect consumer rights.

[0003] With the development of machine learning technology, classification methods based on neural networks have shown powerful capabilities in pattern recognition and classification problems. Back propagation neural networks, as a common multi-layer feedforward neural network, have been widely used in the field of food traceability such as tea, shrimp, and corn. However, back propagation neural networks are prone to fall into local optimality during training, and their training efficiency and classification performance are greatly affected by the initial weights and parameter settings, which often leads to poor model effects. Particle swarm optimization, as an intelligent optimization algorithm, simulates the foraging process of bird flocks to perform global search, and has the advantages of simple implementation and fast convergence speed. Particle swarm optimization can effectively overcome the shortcomings of traditional back propagation neural networks that are prone to fall into local optimality, improve the accuracy and robustness of the model, and provide a new idea for the identification of the origin of Tan mutton. Summary of the invention

[0004] The purpose of the present invention is to provide a method for identifying the origin of Tan mutton based on a back propagation neural network, so as to identify the origin of Tan mutton quickly and reliably with a small amount of sampling.

[0005] In order to achieve these purposes and other advantages according to the present invention, a method for identifying the origin of Tan mutton based on a back propagation neural network is provided, comprising: Extract lipids from Tan mutton samples from different origins; The composition and content of lipids in Tan mutton samples from different origins were analyzed using LC-MS / MS system. The random forest model was used to screen the composition and content of lipids in all samples, and the characteristic lipid data of Tan mutton samples from different origins were obtained. The characteristic lipid data of Tan mutton samples from different origins were used to train the pre-built back propagation neural network optimized by particle swarm algorithm. The trained back propagation neural network optimized by particle swarm algorithm was used to identify the origin of Tan mutton of unknown origin.

[0006] Preferably, the method for extracting lipids from a Tan sheep meat sample comprises: Take 20 mg of Tan mutton sample, add 1 mL of lipid extract containing internal standard, vortex for 15 min, add 200 μL of water, vortex for 1 min, then centrifuge at 13500g for 10 min at 4°C, aspirate 200 μL of the supernatant and evaporate it with a vacuum concentrator, dissolve the dried extract in 200 μL of lipid reconstitution solution, vortex for 3 min, and centrifuge at 13500g for 3 min.

[0007] Preferably, the lipid extract containing the internal standard is a solution obtained by mixing methyl tert-butyl ether and methanol in a volume ratio of 3:1; The lipid reconstitution solution is a solution obtained by mixing acetonitrile and isopropanol in a volume ratio of 1:1.

[0008] Preferably, the chromatographic conditions for LC-MS / MS system analysis are: using a C30 chromatographic column for chromatographic separation, the chromatographic column temperature is 45°C, and the lipids are gradient eluted using a binary solvent system, in which: mobile phase A includes an acetonitrile-water solution containing 0.1% by volume formic acid and 10 mM ammonium formate, and the volume ratio of acetonitrile to water is 6:4; mobile phase B includes an acetonitrile-isopropanol solution containing 0.1% by volume formic acid and 10 mM ammonium formate, and the volume ratio of acetonitrile to isopropanol is 1:9, and the gradient The elution flow rate was 0.35 mL / min, and the gradient elution parameters included: from 0 to 15.5 min, mobile phase A decreased from 80% to 5%, and mobile phase B increased from 20% to 95%; from 15.5 to 17.3 min, mobile phase A was 5%, and mobile phase B was 95%; from 17.3 to 17.5 min, mobile phase A increased from 5% to 80%, and mobile phase B decreased from 95% to 20%; from 17.5 to 20 min, mobile phase A was 80%, and mobile phase B was 20%.

[0009] Preferably, the mass spectrometry conditions for LC-MS / MS system analysis are: using ESI-QTRAP-MS mode, electrospray ion source temperature of 500°C, mass spectrometry voltage of 5500V in positive ion mode, and mass spectrometry voltage of -4500V in negative ion mode.

[0010] Preferably, the data acquisition conditions for LC-MS / MS system analysis are: extracting raw peak intensity data, the minimum intensity of chromatographic peak filtering is 1000 cps, the minimum signal-to-noise ratio is 5, the retention time deviation is <0.2min, based on the QC sample peak area, deleting substances with a detection rate less than 50% and a CV greater than 0.3.

[0011] Preferably, the method of using the random forest model to screen the composition and content of lipids in all beach mutton samples from different origins includes: setting the number of classification trees to 1000, randomly extracting one-fifth of the data as out-of-bag data, and repeatedly constructing each model 10 times to avoid randomness in the results, evaluating the model performance by the average OOB error rate, ranking the factors by the average variable importance, and gradually eliminating the factors with the lowest importance of 20%.

[0012] Preferably, a particle swarm algorithm is used to optimize the back propagation neural network, and the parameters are set as follows: the number of population updates of the particle swarm algorithm is 30, the population size is 5, the maximum speed is 1.0, the minimum speed is -1.0, the optimization variable boundary is -2.0 to 2.0, the number of input layer nodes of the back propagation neural network is 11, the number of hidden layer nodes is 6, and the number of output layer nodes is 2.

[0013] The present invention includes at least the following beneficial effects: a feasible method for distinguishing the origins of Tan mutton from different regions is constructed using lipidomics and machine learning algorithms. Extensive targeted lipidomics analysis was performed on meat samples from different regions, and the most effective characteristic lipids were screened out by the random forest algorithm, and the accuracy of the back propagation (BP) neural network and two optimized BP neural networks (based on genetic algorithm optimization (GA) and particle swarm algorithm optimization (PSO)) in distinguishing the origins of Tan mutton was explored. The results show that random forest can be used as a biomarker screening method to achieve effective screening of characteristic lipids, and the back propagation neural network optimized by the particle swarm algorithm can achieve accurate identification of the origins of Tan mutton from different regions, which helps to provide reliable technical support for distinguishing meat from different sources.

[0014] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the method for identifying the origin of Tan mutton based on back propagation neural network according to an embodiment of the present invention; Figure 2a This is the prediction accuracy graph of the BP neural network using the training set; Figure 2b This is the prediction accuracy graph of the BP neural network using the test set; Figure 3a The confusion matrix diagram of the BP neural network using the training set; Figure 3b The confusion matrix diagram of the BP neural network using the test set; Figure 4a This is the prediction accuracy graph of the GA-BP neural network using the training set; Figure 4b This is the prediction accuracy graph of the GA-BP neural network using the test set; Figure 5a Confusion matrix diagram of the GA-BP neural network using the training set; Figure 5b Confusion matrix diagram of the GA-BP neural network using the test set; Figure 6a This is the prediction accuracy graph of the PSO-BP neural network using the training set; Figure 6b This is the prediction accuracy graph of the PSO-BP neural network using the test set; Figure 7a Confusion matrix diagram of the training set used for the PSO-BP neural network; Figure 7b Confusion matrix diagram of the PSO-BP neural network using the test set. DETAILED DESCRIPTION

[0016] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0017] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials are commercially available unless otherwise specified; in the description of the present invention, the terms "lateral", "longitudinal", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0018] The applicant collected samples of Tan mutton produced in Yanchi County of Ningxia Hui Autonomous Region (NX), Jingyuan County of Gansu Province (GS) and Etuoke Front Banner of Inner Mongolia Autonomous Region (IM) for testing.

[0019] The lipid standards used in the experimental tests were 12:0 Lyso PC, Cer (d18:1 / 4:0), PC (13:0 / 13:0), DG (12:0 / 12:0), and TG (17:0 / 17:0 / 17:0).

[0020] The chemicals used in the experimental tests are: chromatography grade acetonitrile, methanol, isopropanol and methyl tert-butyl ether, formic acid, dichloromethane. Mass spectrometry grade ammonium formate.

[0021] The ultrapure water used in the experimental tests was produced by a Milli-Q system.

[0022] like Figure 1 As shown, the method for identifying the origin of Tan mutton based on a back propagation neural network according to an embodiment of the present invention comprises: S1. Extract lipids from mutton samples from different origins; Specifically, the method for extracting lipids from a Tan mutton sample comprises: Take 20 mg of Tan mutton sample, add 1 mL of lipid extract containing internal standard, vortex for 15 min, add 200 μL of water, vortex for 1 min, then centrifuge at 13500g for 10 min at 4°C, aspirate 200 μL of the supernatant and evaporate it with a vacuum concentrator, dissolve the dried extract in 200 μL of lipid reconstitution solution, vortex for 3 min, and centrifuge at 13500g for 3 min.

[0023] The lipid extract containing the internal standard was a solution obtained by mixing methyl tert-butyl ether and methanol in a volume ratio of 3:1; The lipid reconstitution solution is a solution obtained by mixing acetonitrile and isopropanol in a volume ratio of 1:1.

[0024] S2. Analyze the composition and content of lipids in mutton samples from different origins using LC-MS / MS system; Specifically, the chromatographic conditions for the LC-MS / MS system analysis are: a Thermo Accucore™ C30 column with a specification of id2.1×100mm, 2.6μm, mobile phase A is an acetonitrile-water solution containing 0.1% volume fraction formic acid and 10mM ammonium formate, and the volume ratio of acetonitrile to water is 6:4, and mobile phase B is an acetonitrile-isopropanol solution containing 0.1% volume fraction formic acid and 10mM ammonium formate, and the volume ratio of acetonitrile to isopropanol is 1:9, and the mobile phase elution gradient is: 0min: mobile phase A is 80%, mobile phase B is 20%; 2 min: mobile phase A 70%, mobile phase B 30%; 4 min: mobile phase A 40%, mobile phase B 60%; 9min: mobile phase A: 15%, mobile phase B: 85%; 14min: mobile phase A: 10%, mobile phase B: 90%; 15.5min: mobile phase A: 5%, mobile phase B: 95%; 17.3min: mobile phase A: 5%, mobile phase B: 95%; 17.5min: mobile phase A: 80%, mobile phase B: 20%; 20min: mobile phase A: 80%, mobile phase B: 20%; Flow rate: 0.35 mL / min, column temperature: 45 °C.

[0025] The mass spectrometry conditions for LC-MS / MS system analysis were: ESI-QTRAP-MS mode, electrospray ion source temperature of 500°C, mass spectrometry voltage of 5500 V in positive ion mode, and mass spectrometry voltage of -4500 V in negative ion mode.

[0026] The data acquisition conditions for the LC-MS / MS system analysis were as follows: the raw peak intensity data (*.wiff) in the Analyst® software were extracted using MultiQuant 3.0.3 software, the minimum intensity for chromatographic peak filtering was 1000 cps, the minimum signal-to-noise ratio was 5, the retention time deviation was <0.2 min, and substances with a detection rate of less than 50% and a CV greater than 0.3 were deleted based on the peak area of ​​the QC samples.

[0027] S3. The random forest model was used to screen the composition and content of lipids in all samples to obtain the characteristic lipid data of Tan mutton samples from different origins; Specifically, the method of using the random forest model to screen the composition and content of lipids in all samples of Tan mutton from different origins includes: setting the number of classification trees to 1000, randomly extracting one-fifth of the data as out-of-bag data, and repeating each model 10 times to avoid randomness of the results. The model performance is evaluated by the average OOB error rate, and the factors are ranked by the average variable importance. The factors with the lowest importance of 20% are gradually eliminated. The screening results are shown in Table 1: Table 1 Feature lipid data screening process Number of variables Out-of-bag error rate (%) Average OOB error rate (%) 1080 23.88 7.46 864 20.00 2.69 691 17.22 5.80 552 9.44 3.56 441 6.11 3.88 358 1.11 2.22 286 2.78 2.78 229 5.00 1.67 183 5.00 1.67 146 2.22 2.72 117 1.11 2.21 94 0 0 75 0 0 60 0 0 48 0 0 38 0 0 30 0 0 24 0 0 19 0 0 15 0 0 12 0 0 10 0 0 8 6.11 3.88 6 2.78 2.78 5 2.78 2.78 4 3.34 2.72 3 4.45 2.22 2 7.22 2.54 As can be seen from Table 1, the random forest model was used for screening. From the stepwise screening process, it can be seen that as the number of lipids decreased, the classification error of the model showed an overall trend of first decreasing and then increasing. The random forest model achieved a classification error of 0 when the number of lipids was 94, indicating that the 94 lipids were an important factor in distinguishing the Tan mutton samples from the three regions. In the ongoing lipid elimination process, the number of characteristic lipids gradually decreased from 94 to 10, and the classification error of the random forest model remained at 0. After further reducing the number of lipids, the model produced classification errors. Therefore, the 10 characteristic lipids finally screened out (as shown in Table 2) are the key to distinguishing the Tan mutton samples from the three regions.

[0028] Table 2 Characteristic lipid data for identifying the origin of Tan sheep Lipids Molecular formula Relative molecular weight (Da) Ionization Model Lipid Class Carnitine C16:0 <![CDATA[C 23 H 45 NO4]]> 399.33 [M+H]+ Fatty Acyls Carnitine C17:0 <![CDATA[C 24 H 47 NO4]]> 413.35 [M+H]+ Fatty Acyls Carnitine C18:0 <![CDATA[C 25 H 49 NO4]]> 427.36 [M+H]+ Fatty Acyls Carnitine C3:0 <![CDATA[C 10 H 19 NO4]]> 217.13 [M+H]+ Fatty Acyls Carnitine C5:1 <![CDATA[C 12 H 21 NO4]]> 243.14 [M+H]+ Fatty Acyls PC(16:0_16:1) <![CDATA[C 40 H 78 NO8P]]> 731.54 [M+COOH]- Glycerophospholipids PC(18:1_24:1) <![CDATA[C 50 H 96 NO8P]]> 869.68 [M+COOH]- Glycerophospholipids PG(16:0_20:0) <![CDATA[C 42 H 83 O 10 P]]> 778.57 [MH]- Glycerophospholipids PI(18:0_18:1) <![CDATA[C 45 H 85 O 13 P]]> 864.57 [MH]- Glycerophospholipids PS(18:0_16:1) <![CDATA[C 40 H 76 NO 10 P]]> 761.52 [MH]- Glycerophospholipids S4, using characteristic lipid data of Tan mutton samples from different origins to train the pre-built back propagation neural network optimized by particle swarm algorithm; Specifically, the particle swarm algorithm is used to optimize the back propagation neural network, and the parameters are set as follows: the number of population updates of the particle swarm algorithm is 30, the population size is 5, the maximum speed is 1.0, the minimum speed is -1.0, the optimization variable boundary is -2.0 to 2.0, the number of input layer nodes of the back propagation neural network is 11, the number of hidden layer nodes is 6, and the number of output layer nodes is 2.

[0029] S5. Use the trained back propagation neural network optimized by particle swarm algorithm to identify the origin of Tan mutton of unknown origin.

[0030] In order to further study the advantages and disadvantages of different machine learning classification methods for the identification of Yanchi Beach mutton, this application also used an unoptimized back propagation neural network (BP) and a back propagation neural network based on genetic algorithm optimization (GA) for comparison.

[0031] Among them, the parameters of the unoptimized back propagation neural network (BP) are set as follows: the number of input layer nodes is 11, the number of hidden layer nodes is 6, and the number of output layer nodes is 2.

[0032] The parameters of the genetic algorithm (GA) are set as follows: genetic generation is 30, population size is 5, the number of optimization parameters is “S = number of input nodes * number of hidden layer nodes + number of hidden layer nodes * number of output layer nodes + number of hidden layer nodes + number of output layer nodes”, and the optimization variable boundary is -2.0 to 2.0.

[0033] The above three neural networks (BP, GA-BP, PSO-BP) were used to discriminate whether the lamb samples came from the salt pond based on the characteristic lipid data of Tan sheep meat from different regions. The analysis process of all methods was carried out by MATLABR2023b. The performance of the model was evaluated by the obtained confusion matrix. The characteristic lipid data were randomly divided into training set and test set at a ratio of 4:1.

[0034] The test results are shown in Figures 2 to 7: Figure 2a This is the prediction accuracy graph of the BP neural network using the training set. Figure 2b This is the prediction accuracy graph of the BP neural network using the test set. Figure 3a This is the confusion matrix diagram of the BP neural network using the training set. Figure 3b This is the confusion matrix diagram of the BP neural network using the test set. Figure 4a This is the prediction accuracy graph of the GA-BP neural network using the training set. Figure 4b This is the prediction accuracy graph of the GA-BP neural network using the test set. Figure 5a This is the confusion matrix diagram of the GA-BP neural network using the training set. Figure 5b This is the confusion matrix diagram of the GA-BP neural network using the test set. Figure 6aThis is the prediction accuracy graph of the PSO-BP neural network using the training set. Figure 6b This is the prediction accuracy graph of the PSO-BP neural network using the test set. Figure 7a The confusion matrix diagram of the PSO-BP neural network using the training set, Figure 7b Confusion matrix diagram of the PSO-BP neural network using the test set.

[0035] As shown in Figures 2 to 7, the prediction performance of the BP neural network for the samples of Tan mutton from different regions is 78.57% for the training set and 75.00% for the test set. The BP neural network performed poorly in the discrimination of Tan mutton from different regions. The optimized BP neural network has improved the prediction performance to varying degrees compared with the single BP neural network. The prediction accuracy of the GA-BP neural network model training set has increased to 100.00%, but its prediction accuracy of the test set is 75%, which still cannot achieve accurate discrimination of Tan mutton from different regions. The PSO-BP neural network model showed amazing prediction performance, with the prediction accuracy of both the training set and the test set reaching 100%.

[0036] Compared with the single BP neural network searching for optimal parameters, the advantage of the PSO algorithm in avoiding falling into the local optimal solution during the search process and then finding the solution closest to the global optimal solution may be the key to the significant improvement of the prediction performance of the BP neural network after the PSO algorithm is optimized. In comparison with the GA-BP model, PSO-BP showed better performance. This may be due to the difference in the optimization methods of the two algorithms. In contrast to the GA algorithm, the PSO algorithm has the characteristic of memory, which enables all particles to save good solution knowledge. Secondly, during the optimization process of the GA algorithm, each chromosome shares information with each other, while the search process of the PSO algorithm is independent, which may also be the key to the performance difference between the two.

[0037] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for identifying the origin of Tan mutton based on back propagation neural network, characterized in that: include: Extract lipids from Tan mutton samples from different origins; The composition and content of lipids in Tan mutton samples from different origins were analyzed using LC-MS / MS system. The random forest model was used to screen the composition and content of lipids in all samples, and the characteristic lipid data of Tan mutton samples from different origins were obtained. The characteristic lipid data of Tan mutton samples from different origins were used to train the pre-built back propagation neural network optimized by particle swarm algorithm. The trained back propagation neural network optimized by particle swarm algorithm was used to identify the origin of Tan mutton of unknown origin.

2. The method for identifying the origin of Tan mutton based on back propagation neural network according to claim 1, characterized in that: The method for extracting lipids from Tan mutton samples includes: Take 20 mg of Tan mutton sample, add 1 mL of lipid extract containing internal standard, vortex for 15 min, add 200 μL of water, vortex for 1 min, then centrifuge at 13500g for 10 min at 4°C, aspirate 200 μL of the supernatant and evaporate it with a vacuum concentrator, dissolve the dried extract in 200 μL of lipid reconstitution solution, vortex for 3 min, and centrifuge at 13500g for 3 min.

3. The method for identifying the origin of Tan mutton based on back propagation neural network as claimed in claim 2, characterized in that: The lipid extract containing the internal standard was a solution obtained by mixing methyl tert-butyl ether and methanol in a volume ratio of 3:1; The lipid reconstitution solution is a solution obtained by mixing acetonitrile and isopropanol in a volume ratio of 1:

1.

4. The method for identifying the origin of Tan mutton based on back propagation neural network according to claim 1, characterized in that: The chromatographic conditions for LC-MS / MS system analysis were as follows: a C30 column was used for chromatographic separation, the column temperature was 45°C, and the lipids were gradient eluted using a binary solvent system. In the binary solvent system, mobile phase A included an acetonitrile-water solution containing 0.1% by volume formic acid and 10 mM ammonium formate, with a volume ratio of acetonitrile to water of 6:4; mobile phase B included an acetonitrile-isopropanol solution containing 0.1% by volume formic acid and 10 mM ammonium formate, with a volume ratio of acetonitrile to isopropanol of 1:9, and a gradient of 1:

4. The elution flow rate was 0.35 mL / min, and the gradient elution parameters included: from 0 to 15.5 min, mobile phase A decreased from 80% to 5%, and mobile phase B increased from 20% to 95%; from 15.5 to 17.3 min, mobile phase A was 5%, and mobile phase B was 95%; from 17.3 to 17.5 min, mobile phase A increased from 5% to 80%, and mobile phase B decreased from 95% to 20%; from 17.5 to 20 min, mobile phase A was 80%, and mobile phase B was 20%.

5. The method for identifying the origin of Tan mutton based on back propagation neural network according to claim 4, characterized in that: The mass spectrometry conditions for LC-MS / MS system analysis were: ESI-QTRAP-MS mode, electrospray ion source temperature of 500°C, mass spectrometry voltage of 5500 V in positive ion mode, and mass spectrometry voltage of -4500 V in negative ion mode.

6. The method for identifying the origin of Tan mutton based on back propagation neural network according to claim 5, characterized in that: The data acquisition conditions for LC-MS / MS system analysis were as follows: extraction of raw peak intensity data, minimum intensity of chromatographic peak filtering of 1000 cps, minimum signal-to-noise ratio of 5, retention time deviation <0.2 min, and deletion of substances with a detection rate less than 50% and a CV greater than 0.3 based on the peak area of ​​QC samples.

7. The method for identifying the origin of Tan mutton based on back propagation neural network according to claim 1, characterized in that: The method of using random forest model to screen the composition and content of lipids in all samples of beach mutton from different origins includes: setting the number of classification trees to 1000, randomly selecting one-fifth of the data as out-of-bag data, and repeating each model 10 times to avoid randomness of the results. The model performance is evaluated by the average OOB error rate, and the factors are ranked by the average variable importance, and the factors with the lowest importance of 20% are gradually eliminated.

8. The method for identifying the origin of Tan mutton based on back propagation neural network according to claim 1, characterized in that: The particle swarm algorithm is used to optimize the back propagation neural network. The parameters are set as follows: the number of population updates of the particle swarm algorithm is 30, the population size is 5, the maximum speed is 1.0, the minimum speed is -1.0, the optimization variable boundary is -2.0 to 2.0, the number of input layer nodes of the back propagation neural network is 11, the number of hidden layer nodes is 6, and the number of output layer nodes is 2.

Images