Wheat bread volume prediction method and system based on near infrared reflectivity and bidirectional integrated learning

Abstract

The invention belongs to the technical field of wheat bread volume prediction, and relates to a wheat bread volume prediction method and system based on near-infrared reflectivity and bidirectional integrated learning, and the method comprises the steps: obtaining near-infrared spectrum data after wheat is ground into flour and a bread volume after the same batch of flour is made into bread; generating a training set and a verification set according to the near infrared spectrum data and the corresponding bread volume; inputting the training set into a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm, and training the wheat bread volume prediction model; after to-be-detected wheat is ground into flour, near infrared spectrum detection is carried out, near infrared spectrum data of the to-be-detected wheat is input into the trained wheat bread volume prediction model, and a wheat bread volume prediction result is generated. According to the method, the accuracy of wheat bread volume prediction is improved, information related to the wheat bread volume can be accurately estimated in early generation, and powerful support is provided for wheat quality breeding.

Description

Technical Field

[0001] The invention relates to a wheat bread volume prediction method and system based on near-infrared reflectivity and bidirectional integrated learning, belonging to the technical field of wheat bread volume prediction. Background Art

[0002] Wheat bread volume is an important quality index of wheat and a key indicator for measuring bread quality. Although it has been reported to be related to a series of quality parameters, such as protein content, different types of glutenin and alcohol-soluble protein subunits, SDS sedimentation value, and rheological properties of dough, these parameters are not sufficient to predict the overall quality of bread. At present, routine wheat quality analysis in the laboratory is still required to determine the bread quality of wheat. However, this method is labor-intensive and time-consuming in large-scale breeding, such as evaluating the bread volume of thousands of breeding materials, and often requires experienced laboratory technicians. In addition, it also requires a large number of samples, which can usually only be met in high-generation breeding materials, making it impossible to evaluate based on laboratory processing quality in the early generations. Summary of the invention

[0003] In view of the above problems, the purpose of the present invention is to provide a method and system for predicting the volume of wheat bread based on near-infrared reflectivity and two-way integrated learning. By combining near-infrared spectroscopy (NIRS) technology with a machine learning algorithm, it is possible to more economically and quickly accurately estimate information related to the volume of wheat bread in an early generation.

[0004] To achieve the above-mentioned purpose, the present invention proposes the following technical scheme: a method for predicting the volume of wheat bread based on near-infrared reflectivity and bidirectional ensemble learning, comprising the following steps: obtaining near-infrared spectral data of wheat after being ground into flour and the volume of bread made from the same batch of flour; generating a training set and a validation set according to the near-infrared spectral data and the corresponding bread volume; inputting the training set into a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm to train the training set; after the wheat to be tested is ground into flour, near-infrared spectral detection is performed, and the near-infrared spectral data of the wheat to be tested is input into the trained wheat bread volume prediction model to generate a wheat bread volume prediction result.

[0005] Furthermore, the method for obtaining near-infrared spectral data of wheat after it is ground into flour includes: using a Fourier infrared spectrometer to collect the near-infrared spectrum of the wheat flour sample, repeating the test for each wheat flour sample several times, and taking the average value of the multiple tests.

[0006] Furthermore, the method for generating the wheat bread volume prediction model is as follows: extracting a feature subset from the training set, training an integrated regression model generated by a stacking algorithm based on the feature subset, repeating the above steps until a preset number of cycles is reached, and then calculating the predicted volume of the wheat bread according to the R2 The first several stacking models are retained, and the training set is predicted by the retained stacking models, and the results are averaged to obtain the final prediction value.

[0007] Furthermore, the training set includes m features and n samples; the dimension of the feature subset extracted from the training set is n×(a×m), and the optimal value of a is determined by the Boruta algorithm.

[0008] Furthermore, the a is preferably 30%.

[0009] Furthermore, the stacking model includes a plurality of different learners, and a heterogeneous integration method is adopted to train the plurality of different learners in parallel, and the learners are integrated through a meta-model to generate a final wheat bread volume prediction result.

[0010] Furthermore, the stacked model includes two levels. The first level includes parallel PLSR (partial least squares regression), RR (ridge regression) and MLR (multiple linear regression) models. The second level integrates the prediction results of the first level into a new matrix, uses the MLR model as a secondary learner, and inputs the new matrix into the secondary learner to generate the wheat bread volume prediction results.

[0011] Furthermore, in the first layer, the data in the training set are respectively input into parallel PLSR, RR and MLR models to generate prediction results of the PLSR, RR and MLR models respectively; the data in the prediction set are respectively input into parallel PLSR, RR and MLR models to generate prediction results of the PLSR, RR and MLR models respectively; the prediction results of each model of the data in the training set are integrated with the prediction results of each model of the data in the prediction set in the second layer to generate the new matrix.

[0012] Further, the top 10% of the stacked models are retained.

[0013] The present invention also discloses a wheat bread volume prediction system based on near-infrared reflectivity and bidirectional ensemble learning, comprising: a data acquisition module, used to acquire near-infrared spectrum data of wheat after being ground into flour and the volume of bread after the same batch of flour is made into bread; a training set and a validation set generation module, used to generate a training set and a validation set according to the near-infrared spectrum data and the corresponding bread volume; a model training module, used to input the training set into a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm to train the model; a model output module, used to perform near-infrared spectrum detection on the wheat to be tested after being ground into flour, input the near-infrared spectrum data of the wheat to be tested into the trained wheat bread volume prediction model, and generate a wheat bread volume prediction result.

[0014] The technical solution of the present invention has at least the following technical effects or advantages: the present invention constructs a rapid estimation model of wheat bread volume based on near-infrared reflectivity by associating near-infrared reflectivity with a bidirectional integrated learning model, thereby improving the accuracy of wheat bread volume prediction, and can accurately estimate information related to wheat bread volume in the early generation, providing strong support for wheat quality breeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a schematic diagram of a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm in one embodiment of the present invention;

[0016] Figure 2 is a distribution diagram of a bread volume dataset in a training set and a validation set in one embodiment of the present invention, Figure 2 (a) is the distribution diagram of the training set. Figure 2 (b) is the distribution diagram of the validation set;

[0017] Figure 3 is a diagram showing prediction results of a wheat bread volume prediction model according to an embodiment of the present invention;

[0018] Figure 4 This is a diagram of prediction results of a wheat bread volume prediction model combined with the best 10% in one embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail through specific embodiments. However, it should be understood that the provision of specific embodiments is only for a better understanding of the present invention, and they should not be understood as limitations of the present invention. In the description of the present invention, it should be understood that the terms used are only for the purpose of description and cannot be understood as indicating or implying relative importance.

[0020] In order to solve the problems existing in the prior art, such as the need for a large number of wheat samples, reliance on manual experience, time-consuming and labor-intensive, and the fact that samples can usually only be found in high-generation breeding materials, the present invention proposes a wheat bread volume prediction method and system based on near-infrared reflectivity and bidirectional ensemble learning, constructs a near-infrared data set of bread volume containing 5,003 representative wheat samples, and establishes a wheat bread volume prediction model based on a bidirectional ensemble learning (T10SLRE) model, which can quickly and accurately evaluate the bread volume. This method can be widely used in traditional wheat quality evaluation links, and can enable flour mills, bread bakeries, breeders and other related personnel to conveniently evaluate the wheat bread volume without the need for a special research institution. The scheme of the present invention is described in detail below through embodiments in conjunction with the accompanying drawings.

[0021] Embodiment 1

[0022] This embodiment discloses a method for predicting the volume of wheat bread based on near infrared reflectivity and bidirectional ensemble learning. Figure 1 As shown, the following steps are included:

[0023] S1 obtains the near infrared spectrum data of wheat after it is ground into flour and the volume of bread made from the same batch of flour. Determination of bread volume.

[0024] All the collected wheat kernels were first tested for their integrity by falling number test. Then they were ground using Buhler mill according to AACC approved method 26-21.02. The flour obtained was stabilized for more than 15 days at room temperature before further analysis. Bread making was carried out according to the baking method in method AACCI 10-10.03. The bread volume of each material was measured after the standard baking test according to AACCI method 10-05.01 to obtain the wheat bread volume.

[0025] The method of obtaining near infrared spectroscopy data of wheat after grinding into flour includes: using Fourier transform infrared spectrometer FT-NIR (Antaris TM II) Collect near infrared spectra of wheat flour samples, repeat the test several times for each wheat flour sample, and take the average value of multiple tests. In this embodiment, when using a Fourier transform infrared spectrometer, the Fourier transform infrared spectrometer needs to be preheated for 30 minutes before each spectral data collection to reach a stable state. The scanning range is set to 10,000-4,000cm -1 , the spectral resolution is about 4cm -1 The absorbance values ​​were recorded as log 1 / R, where R is the sample reflectance. All spectral analyses were performed in a room at 24 ± 1 °C.

[0026] S2 generates training and validation sets based on near-infrared spectral data and corresponding bread volumes.

[0027] The distribution diagram of the bread volume dataset in the training set and the validation set is as follows: Figure 1 As shown, the training set includes m features and n samples. In this embodiment, the training set includes 4002 samples and the validation set includes 1001 samples.

[0028] S3 inputs the training set into the wheat bread volume prediction model generated based on the bidirectional ensemble learning algorithm T10SLRE to train it.

[0029] like Figure 1 As shown, the method for generating the wheat bread volume prediction model is: extracting a feature subset from the training set, the dimension of the extracted feature subset is n×(a×m), the optimal value of a is determined by the Boruta algorithm, and in this embodiment, the importance of all bands is evaluated according to the Boruta algorithm combined with the random forest algorithm, and the accuracy of bands of different proportions is performed in combination with the multivariate linear regression MLR model, and finally it is determined that the 30% band has the best effect. In this embodiment, a is preferably 30%. Based on the feature subset, the integrated regression model generated by the stacking algorithm is trained, and the above steps are repeated until the preset number of cycles is reached. In this embodiment, the number of cycles is 1000 times, but it is not limited to this. The prediction result of the wheat bread volume prediction model obtained after 1000 cycles is as shown in the figure. Figure 3 As shown. According to the R 2 The first several stacking models are retained. In this embodiment, the first 10% of the stacking models are retained. The training set is predicted by the first 10% of the stacking models, and the results are averaged to obtain the final prediction value. The multi-sampling ensemble method of bidirectional ensemble learning is used to overcome the problem of overfitting of a single model. The prediction results of the final wheat bread volume prediction model are shown in Figure 1. Figure 4 shown. Figure 4 The results show the fitting effect of the wheat bread volume prediction model, and the prediction accuracy of the test set reached R 2 =0.74. The main reason is that the T10SLRE algorithm effectively selects the best spectral bands and reduces the influence of noise and outliers. In addition, the use of low-dimensional spectral bands (30%) reduces the model complexity and enhances the robustness of linear regression. Finally, the T10SLRE algorithm avoids the need for spectral preprocessing and band screening when building the near-infrared model by randomly selecting 30% of the spectral bands and combining the stacking and averaging of the first 10% model results, thereby achieving better prediction capabilities.

[0030] The bidirectional ensemble learning algorithm T10SLRE has demonstrated strong fitting ability in wheat bread volume estimation. In addition, compared with traditional methods, which require near-infrared spectral preprocessing and feature band screening to achieve better prediction capabilities, the bidirectional ensemble learning algorithm T10SLRE can directly analyze near-infrared data, eliminating the steps of preprocessing and feature band screening, and has greatly improved the adaptability and generalization ability of the model.

[0031] The stacking model includes multiple different learners. It adopts a heterogeneous ensemble method to train multiple different learners in parallel, integrate the learners through a meta-model, and generate the final wheat bread volume prediction result.

[0032] The stacking model includes two levels. The first level includes parallel PLSR, RR and MLR models. In the first level, the data in the training set are input into the parallel PLSR, RR and MLR models to generate the prediction results of the PLSR, RR and MLR models respectively. The data in the prediction set are input into the parallel PLSR, RR and MLR models to generate the prediction results of the PLSR, RR and MLR models respectively. The prediction results of each model of the data in the training set are integrated with the prediction results of each model of the data in the prediction set in the second level to generate a new matrix. The second level integrates the prediction results of the first level into a new matrix, uses the MLR model as the secondary learner, and inputs the new matrix into the secondary learner to generate the prediction results of wheat bread volume.

[0033] After S4 grinds the wheat to be tested into flour, it performs near-infrared spectroscopy detection, inputs the near-infrared spectral data of the wheat to be tested into a trained wheat bread volume prediction model, and generates a wheat bread volume prediction result.

[0034] The accuracy of the wheat bread volume prediction model is expressed by the coefficient of determination R 2 And root mean square error RMSE evaluation. Coefficient of determination R 2 The calculation formula for the root mean square error RMSE evaluation is:

[0035]

[0036] Among them, y i is the actual measured volume of wheat bread, is the predicted wheat bread volume, is the average volume of wheat bread actually measured; R 2 The larger it is, the smaller the RMSE is, indicating that the model has stronger predictive ability.

[0037] Embodiment 2

[0038] Based on the same inventive concept, the present invention also discloses a wheat bread volume prediction system based on near infrared reflectivity and bidirectional integrated learning, comprising:

[0039] A data acquisition module, used to obtain near infrared spectrum data of wheat after it is ground into flour and the volume of bread after the same batch of flour is made into bread;

[0040] A training set and a validation set generation module, used to generate a training set and a validation set according to near infrared spectral data and corresponding bread volumes;

[0041] A model training module, used for inputting a training set into a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm to train the model;

[0042] The model output module is used to grind the wheat to be tested into flour, perform near-infrared spectroscopy detection, input the near-infrared spectral data of the wheat to be tested into the trained wheat bread volume prediction model, and generate the wheat bread volume prediction result.

[0043] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0044] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0045] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0046] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not deviate from the spirit and scope of the present invention should be included in the protection scope of the claims of the present invention. The above content is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A wheat bread volume prediction method based on near infrared reflectivity and bidirectional ensemble learning, characterized in that: The following steps are involved: Obtain near infrared spectral data of wheat flour after grinding and the volume of bread made from the same batch of flour; Generate a training set and a validation set according to the near infrared spectral data and the corresponding bread volume; Inputting the training set into a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm to train the model; After the wheat to be tested is ground into flour, near-infrared spectroscopy detection is performed, and the near-infrared spectral data of the wheat to be tested is input into the trained wheat bread volume prediction model to generate the wheat bread volume prediction result.

2. The wheat bread volume prediction method based on near infrared reflectivity and bidirectional ensemble learning according to claim 1, characterized in that: The method for obtaining near infrared spectrum data of wheat after being ground into flour includes: using a Fourier infrared spectrometer to collect the near infrared spectrum of the wheat flour sample, repeating the test for each wheat flour sample for several times, and taking the average value of the multiple tests.

3. The wheat bread volume prediction method based on near infrared reflectivity and bidirectional ensemble learning as claimed in claim 1, characterized in that: The method for generating the wheat bread volume prediction model is as follows: extracting a feature subset from the training set, training an integrated regression model generated by a stacking algorithm based on the feature subset, repeating the above steps until a preset number of cycles is reached, and then calculating the predicted volume of wheat bread according to the R on the training set. 2 The first several stacking models are retained, and the training set is predicted by the retained stacking models, and the results are averaged to obtain the final prediction value.

4. The wheat bread volume prediction method based on near infrared reflectivity and bidirectional ensemble learning as claimed in claim 3, characterized in that: The training set includes m features and n samples; the dimension of the feature subset extracted from the training set is n×(a×m), and the optimal value of a is determined by the Boruta algorithm.

5. The wheat bread volume prediction method based on near infrared reflectivity and bidirectional ensemble learning as claimed in claim 4, characterized in that: The a is 30%.

6. The method for predicting wheat bread volume based on near infrared reflectivity and bidirectional ensemble learning as claimed in claim 3, characterized in that: The stacking model includes a plurality of different learners, and adopts a heterogeneous integration method to train the plurality of different learners in parallel, and integrates the learners through a meta-model to generate a final wheat bread volume prediction result.

7. The method for predicting wheat bread volume based on near infrared reflectivity and bidirectional ensemble learning according to claim 6, characterized in that: The stacked model includes two levels. The first level includes parallel PLSR, RR and MLR models. The second level integrates the prediction results of the first level into a new matrix, uses the MLR model as a secondary learner, and inputs the new matrix into the secondary learner to generate the wheat bread volume prediction results.

8. The method for predicting wheat bread volume based on near infrared reflectivity and bidirectional ensemble learning according to claim 7, characterized in that: In the first layer, the data in the training set are respectively input into parallel PLSR, RR and MLR models to generate prediction results of the PLSR, RR and MLR models respectively; the data in the prediction set are respectively input into parallel PLSR, RR and MLR models to generate prediction results of the PLSR, RR and MLR models respectively; the prediction results of each model of the data in the training set are integrated with the prediction results of each model of the data in the prediction set in the second layer to generate the new matrix.

9. The method for predicting wheat bread volume based on near infrared reflectivity and bidirectional ensemble learning according to claim 7, characterized in that: The top 10% of stacked models are retained.

10. A wheat bread volume prediction system based on near infrared reflectivity and bidirectional ensemble learning, characterized in that: include: A data acquisition module, used to obtain near infrared spectrum data of wheat after it is ground into flour and the volume of bread after the same batch of flour is made into bread; A training set and a validation set generation module, used to generate a training set and a validation set according to the near infrared spectrum data and the corresponding bread volume; A model training module, used for inputting the training set into a wheat bread volume prediction model generated based on a bidirectional ensemble learning algorithm to train the model; The model output module is used to grind the wheat to be tested into flour, perform near-infrared spectroscopy detection, input the near-infrared spectral data of the wheat to be tested into the trained wheat bread volume prediction model, and generate the wheat bread volume prediction result.

Images