CN113378383B - Food supply chain hazard prediction method and device - Google Patents

Food supply chain hazard prediction method and device Download PDF

Info

Publication number
CN113378383B
CN113378383B CN202110647468.XA CN202110647468A CN113378383B CN 113378383 B CN113378383 B CN 113378383B CN 202110647468 A CN202110647468 A CN 202110647468A CN 113378383 B CN113378383 B CN 113378383B
Authority
CN
China
Prior art keywords
prediction
prediction model
super
supply chain
parameters
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110647468.XA
Other languages
Chinese (zh)
Other versions
CN113378383A (en
Inventor
金学波
张佳帅
张家辉
苏婷立
白玉廷
孔建磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Technology and Business University
Original Assignee
Beijing Technology and Business University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Technology and Business University filed Critical Beijing Technology and Business University
Priority to CN202110647468.XA priority Critical patent/CN113378383B/en
Publication of CN113378383A publication Critical patent/CN113378383A/en
Application granted granted Critical
Publication of CN113378383B publication Critical patent/CN113378383B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • Computer Hardware Design (AREA)
  • Animal Husbandry (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Educational Administration (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Agronomy & Crop Science (AREA)
  • Evolutionary Computation (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Mining & Mineral Resources (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a food supply chain hazard prediction method, which comprises the following steps: defining a noise smoothing loss function by adopting a regularization method; constructing a prediction model combining a GRU sub-predictor and the noise smoothing loss function, and training the prediction model; and predicting food supply chain hazards according to the trained prediction model to obtain a prediction result. The invention provides a prediction model combining a GRU sub-predictor and a noise smoothing loss function related to regularization, which can reduce the fitting degree of the prediction model to random noise, improve the prediction accuracy and have better effect in the prediction task with larger probability of actually coping with the problems of large noise, measurement error and the like.

Description

Food supply chain hazard prediction method and device
Technical Field
The present disclosure relates to the field of time series prediction, and in particular, to a method and apparatus for predicting a hazard in a food supply chain.
Background
The food supply chain consists of complex links such as raw material suppliers, manufacturers, sellers, consumers and the like, has wide cross-over regional range and has high quality requirements on each link. Once food safety problems occur, not only is the entire supply chain lost, but also a bad social impact is created. Therefore, whether a multilayer food supply chain having low transparency is safe or not has become an important point of attention.
With the development of the internet of things technology, people can obtain more and more information on a food supply chain. Recently, machine learning, particularly a deep learning method, is increasingly used. The recurrent neural network (Recurrent Neural Network, RNN) serves as an important neural network in the field of deep learning, providing a more efficient solution for analysis of sequence data. There are increasing numbers of food supply chain safety prediction systems currently incorporating machine learning or deep neural networks, but these food safety prediction models have a significant potential risk.
The deep learning model can effectively conduct prediction and auxiliary decision making, and is widely applied to time sequence prediction, automatic driving, recommendation and individualization technologies and other related fields. However, in the field of food safety, measurement noise and errors are unavoidable due to the fact that the actual measured food hazard content may be due to sensor performance, incorrect readings, instrument damage, etc. In the training process of the deep learning model, the selection of a loss function is an important part, and can determine the training effect of the model. Due to the strong nonlinearity and strong randomness of the food supply chain hazard data, it is difficult to completely remove the influence of noise in the noisy data analysis stage, which results in that the obtained estimated true value still has noise, the noise in the data is excessively learned in training, the randomness of the noise can lead to the reduction of the prediction performance, and the learning of the noise can also affect the robustness of the model.
Disclosure of Invention
In order to solve one of the above technical problems, the present invention provides a method and an apparatus for predicting hazards in a food supply chain.
An embodiment of the present invention provides a method for predicting a hazard in a food supply chain, the method comprising:
defining a noise smoothing loss function by adopting a regularization method;
constructing a prediction model combining a GRU sub-predictor and the noise smoothing loss function, and training the prediction model;
and predicting food supply chain hazards according to the trained prediction model to obtain a prediction result.
Preferably, the noise smoothing loss function includes two parts of a measured input data fitness represented by an average absolute error between a predicted value and a true value and a measured input data smoothness represented by a norm of a matrix of the measured input data smoothness for every three points.
Preferably, the calculation process of the norm of the matrix for measuring the smoothness degree of every three points in the input data includes:
defining a punishment matrix for punishing smoothness of input data, and calculating to obtain a matrix for measuring smoothness of every three points in the input data according to the punishment matrix and the input data;
and carrying out norm calculation on the matrix measuring the smoothness degree of each three points in the input data to obtain the norm of the matrix measuring the smoothness degree of each three points in the input data.
Preferably, the training process of the prediction model is as follows: and training the prediction model according to the hazard content source data acquired by the Internet of things platform.
Preferably, the method further comprises:
optimizing the super-parameters of the prediction model through a Bayesian optimization algorithm to obtain optimal super-parameters;
training the prediction model according to the optimal super parameters to obtain an optimal prediction model;
and predicting the food supply chain hazard according to the optimal prediction model to obtain a prediction result.
A second aspect of an embodiment of the present invention provides a food supply chain hazard prediction apparatus, the apparatus comprising a processor configured with processor-executable operating instructions to perform operations comprising:
defining a noise smoothing loss function by adopting a regularization method;
constructing a prediction model combining a GRU sub-predictor and the noise smoothing loss function, and training the prediction model;
and predicting food supply chain hazards according to the trained prediction model to obtain a prediction result.
Preferably, the noise smoothing loss function includes two parts of a measured input data fitness represented by an average absolute error between a predicted value and a true value and a measured input data smoothness represented by a norm of a matrix of the measured input data smoothness for every three points.
Preferably, the processor is configured with processor-executable operating instructions to perform the following operations:
defining a punishment matrix for punishing smoothness of input data, and calculating to obtain a matrix for measuring smoothness of every three points in the input data according to the punishment matrix and the input data;
and carrying out norm calculation on the matrix measuring the smoothness degree of each three points in the input data to obtain the norm of the matrix measuring the smoothness degree of each three points in the input data.
Preferably, the processor is configured with processor-executable operating instructions to perform the following operations:
and training the prediction model according to the hazard content source data acquired by the Internet of things platform.
Preferably, the processor is configured with processor-executable operating instructions to perform the following operations:
optimizing the super-parameters of the prediction model through a Bayesian optimization algorithm to obtain optimal super-parameters;
training the prediction model according to the optimal super parameters to obtain an optimal prediction model;
and predicting the food supply chain hazard according to the optimal prediction model to obtain a prediction result.
The beneficial effects of the invention are as follows: the invention provides a prediction model combining a GRU sub-predictor and a noise smoothing loss function related to regularization, which can reduce the fitting degree of the prediction model to random noise, improve the prediction accuracy and have better effect in the prediction task with larger probability of actually coping with the problems of large noise, measurement error and the like.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute an undue limitation to the application. In the drawings:
FIG. 1 is a flowchart of a method for predicting hazard in a food supply chain according to example 1 of the present invention;
FIG. 2 is a schematic diagram of DON concentration, cadmium concentration, and lead concentration for each link provided in the examples;
FIG. 3 is a comparison of the three loss function predictions provided in the examples;
FIG. 4 is a plot of the convergence of the three loss functions provided in the example;
FIG. 5 is a plot of a predicted sample of cadmium metal content for link 11 provided in the examples;
FIG. 6 is a plot of lead metal content prediction samples for segment 11 provided in the examples;
FIG. 7 is a graph of a predicted sample of link 11DON hazard content provided in the examples.
Detailed Description
In order to make the technical solutions and advantages of the embodiments of the present application more apparent, the following detailed description of exemplary embodiments of the present application is given with reference to the accompanying drawings, and it is apparent that the described embodiments are only some of the embodiments of the present application and not exhaustive of all the embodiments. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other.
Example 1
As shown in fig. 1, the present embodiment proposes a food supply chain hazard prediction method, which includes:
s101, defining a noise smoothing loss function by adopting a regularization method;
s102, constructing a prediction model combining a GRU sub-predictor and the noise smoothing loss function, and training the prediction model;
s103, predicting food supply chain hazards according to the trained prediction model to obtain a prediction result.
In particular, regularization is widely used in machine learning and deep learning. The method has the effect of limiting parameters in the model, so that the parameters of the model are not too large, and the possibility of overfitting the model is reduced. In this embodiment, to reduce model overfitting due to over-learning random noise in the supply chain data, the designed noise smoothing loss function (Noise Smoothing Loss, NSL loss function) is formulated as:
wherein y is i Is a predicted value of the current value,is a true value, beta is a regularization term, P y Is a matrix that measures the smoothness of every three points in the data. The calculation formula of the noise smoothing loss function comprises two parts, wherein the first part measures the fitting degree of data and represents a target for minimizing the square sum of residual errors between an actual sequence and a fitting sequence; and the second part is used for measuring the smoothness and representing the requirement on the smoothness of the sequence in the training process. Wherein the fitting degree is represented by average absolute error, and the smoothing degree is calculated by P y Is realized by the norm of (a). Where β is a regularization term, which is considered a trade-off between fitting and smoothing two targets.
P y Is the key to the implementation of the noise smoothing loss function, and a matrix P is first defined during calculation, and the matrix penalizes the smoothness of the data. The P matrix is shown below：
The dimension of the P matrix is determined by the input data, the dimension of P is (T, T) assuming that the length of the input data is T, and then the matrix P is obtained by multiplying the P and the input data to form the matrix y Finally, the norms which can represent the smoothness degree among every three points of the data are obtained through the calculation of the norms of the matrix.
The data is smoothed in the training process, namely, the smoothing is realized in the model training process, so that the model learns the smoothed data in the training process, and the robustness of the model can be further improved.
The GRU is a variant of the long and short term memory network (Long Short Term Memory, LSTM), both belonging to the recurrent neural network. Compared with the LSTM network, the GRU network has the advantages that only the update gate and the reset gate are arranged in the GRU network structure, and the GRU network is simpler and better in effect. In this embodiment, based on the Keras Tensorflow framework, the data is fitted using the GRU, and the optimization objective during training is replaced by the NSL loss function proposed in this embodiment.
GRU algorithm pseudocode:
(1) Normalizing data set θ
(2) Model learning training data
Learn Hbased on θ
return H
The super-parameter selection of the deep learning model directly determines the performance of the model, and in the embodiment, a Bayesian optimization algorithm is realized by adopting a Hyperopt library: optimization based on Sequence Model (SMBO). The optimized super-parameters mainly comprise the number of neurons in the GRU, the Dropout rate, the training times, the batch size and the optimizer.
When the model parameters are determined, the Bayesian optimization method uses a proxy model to fit a real objective function, and actively selects the most potential evaluation point according to the fitting result. It is necessary to define an objective function g (w) and an optimized hyper-parameter space. The objective function represents the minimum objective that needs to be achieved by bayesian optimization, and the present embodiment uses the root mean square error of the model as the objective function to find the model hyper-parameters that produce the best score on this metric.
Where m is the number of input samples, y i (w) is a predicted value of the value,is a predicted value. The bayesian-optimized proxy function can be expressed in the form of an equation as:
wherein w is * And (3) determining optimal parameters for Bayesian optimization, wherein W is an input set of super parameters, and W is a parameter space of the multidimensional super parameters.
Bayesian optimization consists mainly of two steps: the gaussian process is estimated and updated first by step t+1, and then the sampling of the super-parameters is guided by maximizing the proxy function. In the gaussian process, the present embodiment sets the objective function g (w) to follow the following gaussian distribution:
g(w)～GP(μ(w),K(w,w′))
where μ (w) is the mean of g (w), K (w, w ') is the covariance matrix of g (w), and the initial K (w, w') can be expressed as:
in Bayesian optimization, the covariance matrix of the Gaussian process varies with the iterative process, assuming a set of inputs at step t+1The parameter is w t+1 The covariance matrix at this time can be expressed as:
wherein k= [ k (w) t+1 ,w 1 ),k(w t+1 ，w 2 )，...，k(w t+1 ,w t )]In this case, the posterior probability of the objective function can be obtained in this embodiment:
wherein θ is the observed data, μ t+1 (w) is the mean value of the (t+1) th step g (w),is the variance of step g (w) at t+1.
After posterior probability is obtained, the optimal super-parameters are searched by a super-parameter searching method, and the invention uses UCB acquisition functions to complete super-parameter searching:
wherein ζ t+1 Is a constant, S (w|theta t ) For UCB acquisition function, w t+1 Is the selected super parameter of the t+1 step.
The pseudo code of the Bayesian optimization algorithm is as follows:
input: θ is the dataset, g (W) is the RMSE of the model, W is the hyper-parameter space (W ε W), H (w|θ) i ) Is UCB acquisition function, T is the number of super parameters to be selected, and l is the number of sub-sequences of wavelet decomposition.
And (3) outputting: optimum super parameter w * 。
(1) Initializing, θ (l) ←InitSamples(g(w)，θ，l)
(2)
(3) Modeling an objective function g (w), and calculating a posterior probability
(4) Parameter updating, w, using UCB acquisition function * ←arg max H(w|θ i (l) )
(5) Using w * The super-parameters train the model provided by the invention to obtain the prediction y i ←g(w * ) Calculate and update
(6)
(7)endfor
(8)
(9)return w *
The super parameters of the GRU sub-predictors in the training of the prediction model are determined by a Bayesian optimization algorithm. The same data test set was used to compare the gru_nsl with models employing different loss functions, LSTM, and model performance was assessed by calculating root mean square error (Root Mean Square Error, RMSE), mean absolute error (Mean Absolute Error, MAE), pearson correlation coefficient (Pearson correlation coefficient, R). Wherein smaller indices of RMSE and MAE indicate more accurate predictions, and larger values of pearson correlation coefficients indicate tighter fitting of observed values to predicted values. The calculation formula of the evaluation index is as follows:
where m is the number of samples, y is the actual value,for predictive value +.>Mean value of the true values for data, +.>The average value of the prediction results is shown.
Example 2
Corresponding to embodiment 1, this embodiment proposes a food supply chain hazard prediction apparatus comprising a processor configured with processor-executable operating instructions to perform operations of:
defining a noise smoothing loss function by adopting a regularization method;
constructing a prediction model combining a GRU sub-predictor and the noise smoothing loss function, and training the prediction model;
and predicting food supply chain hazards according to the trained prediction model to obtain a prediction result.
Specifically, the specific working principle of the device according to this embodiment may refer to the content described in embodiment 1, and will not be described herein. The embodiment provides a prediction model combining a GRU sub-predictor and a noise smoothing loss function related to regularization, which can reduce the fitting degree of the prediction model to random noise, improve the prediction accuracy and have better effect in the prediction task with larger probability of actually coping with large noise, error measurement and the like.
The prediction process of the prediction method proposed by the present invention and the actual effects presented are further described below by means of two specific examples.
Example 1
Using a built food supply chain hazard prediction model, using the first six links (X 1 ～X 6 ) Five post-hazard content prediction (X) 7 ～X 11 ) And (5) hazard content data. The experiment uses the combination of GRU and average absolute error (Mean Absolute Error, MAE) and mean square error (Mean Square Error, MSE) loss functions as comparison with the model provided by the invention, and the variables such as the control network layer number and the like are kept consistent. The performance of the system was evaluated using three evaluation indexes RMSE, MAE, R.
Firstly, data used in the experiment are described, the data used in the experiment come from a wheat flour supply chain, and the contents of Deoxynivalenol (DON), lead and cadmium hazardous substances in the wheat flour supply chain are respectively collected. The supply chain has 10 links of cleaning, wheat wetting, processing 1, etc., and the raw grain is taken as X 1 Clean wheat is X 2 The wheat is moistened with X 3 Processing 1M core to X 4 Machining a second (2M core) to X 5 Machining a tri (3M core) to X 6 Processing four (4M core) to X 7 Processing five (5M core) to X 8 Machining six (6M cores) to X 9 Packaging as X 10 The warehouse (the detection value of the circulation link in the existing data) is X 11 . And finally, respectively obtaining data of three hazard contents of cadmium metal, lead metal and DON by carrying out spot check on wheat flour in each link. Wherein DON 396 group, lead 1061 group and cadmium 2057 group are respectively in the data formats of (396,11), (1061,11) and (2057,11), and the content of the harmful substances in each link is shown in figure 2.
This comparative experiment uses only cadmium hazard content data. Wherein the partitioning training set format is (1857,6) and the label format is (1857,5); the test set is (200, 6), and the label is (200, 5). The training set label is mainly used for adjusting the weights and the biases of all the neurons of the model in the training process, and the test set label is used for checking the prediction result.
Table 1 shows training results for three loss function combinations of GRU and NSL, MAE, MSE on a cadmium hazard data set. The results show that: the GRU_NSL provided by the invention performs best, and the RMSE reaches 2.4484. The GRU_NSL model is improved by 9.73%, 6.31% and 0.11% on the RMSE, MAE, R three indexes compared with the relatively optimal GRU_MAE model.
TABLE 1
Fig. 3 shows the prediction results of the GRU model trained under three loss functions, respectively, and it can be seen that NSL loss functions have significant advantages. In order to highlight the situation of the NSL training process, fig. 4 shows the convergence situation of the three Loss functions iterating the same number of times in the training process of the GRU model, where the NSL decreases at the fastest speed in the iteration process, and the Loss value is finally stabilized at 0.40, where the stable value is the Loss value after the noise variance is considered, so that the Loss is slightly larger than the Loss of MAE and MSE.
The experimental results show that: in the task of predicting the content of the first 6 link dangers and the last 5 link dangers by using the model, the regularized loss function is combined with the GRU model, so that the effect of improving the prediction performance of the food supply chain dangers prediction model can be achieved.
Example 2:
and predicting the content data of the eleventh-link hazard by using the constructed food supply chain hazard prediction model.
The experiment is carried out based on the content data of cadmium metal, lead metal and DON of the spot check in 11 links of the wheat flour supply chain. Since the warehouse storage link is the last link of food flowing into the market and represents the quality of wheat flour in the market, the experiment is set to predict the result of the last warehouse storage link by using the first 10 links. The GRU_NSL model and the RNN, LSTM, GRU model provided by the invention are used for comparison in experiments, the training set accounts for 80% of the total data, and the rest is used as a test set. The experiment was evaluated using RMSE, MAE, R three indices. According to national standard of food safety, the standard value of cadmium metal in food in agricultural products is not higher than 0.05mg/kg, the standard value of lead metal in agricultural products is not higher than 0.1mg/kg, and DON is not higher than 1000 mug/kg.
Tables 2, 3 and 4 show the experimental results of cadmium data, lead data and DON data, respectively, and all the results are unified in ug/kg. The gru_nsl model performs well in the predictive task of three sets of data, where RMSE reaches 0.2595ug/kg, far less than the upper limit of the cadmium metal content of the national food standard, in a cadmium data application, with a 46.37% improvement in the gru_nsl model based on the RMSE index of the GRU model. In lead data applications, the RMSE of the gru_nsl model reached 0.2700ug/kg, a 50.52% improvement on the RMSE index of the GRU. In DON data application, the GRU_NSL model RMSE reaches 1.5245ug/kg, which is far smaller than the national index, and the model is improved by 70.06% on the basis of the GRU RMSE.
TABLE 2
TABLE 3 Table 3
TABLE 4 Table 4
Fig. 5 to 7 show partial prediction results, wherein 2 sets of prediction results are randomly extracted from the results of three sets of application experiments, and each set of results shows the prediction results of RNN, LST, GRU and the gru_nsl model, wherein the predicted result of the gru_nsl model is closest to the true value.
Therefore, through experimental verification, the GRU_NSL model is better in prediction task, and NSL can further improve the analysis capability of the GRU_NSL model on noise-containing time sequence data. Therefore, the GRU_NSL deep learning unit can effectively analyze noisy time series data, and has good prediction performance on noisy data such as food supply chain dangers and the like.
It will be apparent to those skilled in the art that various modifications and variations can be made in the present application without departing from the spirit or scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims and the equivalents thereof, the present application is intended to cover such modifications and variations.

Claims (4)

1. A method of predicting food supply chain hazards, the method comprising:
defining a noise smoothing loss function by adopting a regularization method, wherein the noise smoothing loss function is as follows:
wherein->Is a predictive value->Is a true value, < >>Is regularized item, +.>Is a matrix for measuring the smoothness of every three points in the data;
constructing a prediction model combining a GRU sub-predictor and the noise smoothing loss function, and training the prediction model;
predicting food supply chain hazards according to the trained prediction model to obtain a prediction result;
the method further comprises the steps of:
optimizing the super-parameters of the prediction model through a Bayesian optimization algorithm to obtain optimal super-parameters;
training the prediction model according to the optimal super parameters to obtain an optimal prediction model;
predicting food supply chain hazards according to the optimal prediction model to obtain a prediction result;
the process for optimizing the super-parameters of the prediction model through the Bayesian optimization algorithm to obtain the optimal super-parameters comprises the following steps:
s1, initializing the weight of a network;
s2, calculating posterior probability of training root mean square error of the prediction model;
s3, performing parameter updating by using a UCB acquisition function, and obtaining super parameters by utilizing super parameter searching；
S4, useThe super-parameters train the prediction model to obtain a prediction result, and the root mean square error is calculated and updated;
s5, repeating the calculation processes from S2 to S4 until all the subsequences are completely calculated, and recording the super parameters at the momentAs an optimal super parameter.
2. The method of claim 1, wherein the calculating of the norms of the matrix measuring the smoothness of every three points in the input data comprises:
defining a punishment matrix for punishing smoothness of input data, and calculating to obtain a matrix for measuring smoothness of every three points in the input data according to the punishment matrix and the input data;
and carrying out norm calculation on the matrix measuring the smoothness degree of each three points in the input data to obtain the norm of the matrix measuring the smoothness degree of each three points in the input data.
3. The method of claim 1, wherein the training the predictive model is: and training the prediction model according to the hazard content source data acquired by the Internet of things platform.
4. A food supply chain hazard prediction device comprising a processor configured with processor-executable operating instructions to perform the food supply chain hazard prediction method of any one of claims 1 to 3.
CN202110647468.XA 2021-06-10 2021-06-10 Food supply chain hazard prediction method and device Active CN113378383B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110647468.XA CN113378383B (en) 2021-06-10 2021-06-10 Food supply chain hazard prediction method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110647468.XA CN113378383B (en) 2021-06-10 2021-06-10 Food supply chain hazard prediction method and device

Publications (2)

Publication Number Publication Date
CN113378383A CN113378383A (en) 2021-09-10
CN113378383B true CN113378383B (en) 2024-02-27

Family

ID=77573591

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110647468.XA Active CN113378383B (en) 2021-06-10 2021-06-10 Food supply chain hazard prediction method and device

Country Status (1)

Country Link
CN (1) CN113378383B (en)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10210860B1 (en) * 2018-07-27 2019-02-19 Deepgram, Inc. Augmented generalized deep learning with special vocabulary
CN109993668A (en) * 2019-04-09 2019-07-09 桂林电子科技大学 A kind of recommending scenery spot method based on gating cycle unit neural network
CN110232480A (en) * 2019-03-01 2019-09-13 电子科技大学 The item recommendation method and model training method realized using the regularization stream of variation
CN110674964A (en) * 2019-05-14 2020-01-10 南京邮电大学 Search prediction system and method based on agricultural traceability information
CN111160667A (en) * 2020-01-02 2020-05-15 北京工商大学 Method and device for improving robustness of food safety prediction model
WO2020143409A1 (en) * 2019-01-07 2020-07-16 阿里巴巴集团控股有限公司 Method and device for predicting business indicators
CN111784159A (en) * 2020-07-01 2020-10-16 深圳市检验检疫科学研究院 Food risk tracing information grading method and device
CN111839713A (en) * 2019-04-24 2020-10-30 上海美杰医疗科技有限公司 Multi-modal tumor ablation probe system and control method thereof
KR20200129058A (en) * 2019-05-07 2020-11-17 한국전자통신연구원 Apparatus and method for speech processing using a densely connected hybrid neural network
CN112015719A (en) * 2020-08-27 2020-12-01 河海大学 Regularization and adaptive genetic algorithm-based hydrological prediction model construction method

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6582429B2 (en) * 2001-07-10 2003-06-24 Cardiac Pacemakers, Inc. Ablation catheter with covered electrodes allowing electrical conduction therethrough
DE102009048312B4 (en) * 2009-07-07 2017-05-11 Erbe Elektromedizin Gmbh Electrosurgical instrument and method for manufacturing an electrosurgical instrument
WO2012126741A2 (en) * 2011-03-18 2012-09-27 Telefonica, S.A. Method for context-aware recommendations based on implicit user feedback
US9204918B2 (en) * 2011-09-28 2015-12-08 RELIGN Corporation Medical ablation system and method of use
US20200348662A1 (en) * 2016-05-09 2020-11-05 Strong Force Iot Portfolio 2016, Llc Platform for facilitating development of intelligence in an industrial internet of things system
US11042796B2 (en) * 2016-11-03 2021-06-22 Salesforce.Com, Inc. Training a joint many-task neural network model using successive regularization
JP2020530159A (en) * 2017-08-02 2020-10-15 ストロング フォース アイオーティ ポートフォリオ ２０１６，エルエルシー Methods and systems for detection of industrial Internet of Things data collection environments using large datasets

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10210860B1 (en) * 2018-07-27 2019-02-19 Deepgram, Inc. Augmented generalized deep learning with special vocabulary
WO2020143409A1 (en) * 2019-01-07 2020-07-16 阿里巴巴集团控股有限公司 Method and device for predicting business indicators
CN110232480A (en) * 2019-03-01 2019-09-13 电子科技大学 The item recommendation method and model training method realized using the regularization stream of variation
CN109993668A (en) * 2019-04-09 2019-07-09 桂林电子科技大学 A kind of recommending scenery spot method based on gating cycle unit neural network
CN111839713A (en) * 2019-04-24 2020-10-30 上海美杰医疗科技有限公司 Multi-modal tumor ablation probe system and control method thereof
KR20200129058A (en) * 2019-05-07 2020-11-17 한국전자통신연구원 Apparatus and method for speech processing using a densely connected hybrid neural network
CN110674964A (en) * 2019-05-14 2020-01-10 南京邮电大学 Search prediction system and method based on agricultural traceability information
CN111160667A (en) * 2020-01-02 2020-05-15 北京工商大学 Method and device for improving robustness of food safety prediction model
CN111784159A (en) * 2020-07-01 2020-10-16 深圳市检验检疫科学研究院 Food risk tracing information grading method and device
CN112015719A (en) * 2020-08-27 2020-12-01 河海大学 Regularization and adaptive genetic algorithm-based hydrological prediction model construction method

Also Published As

Publication number Publication date
CN113378383A (en) 2021-09-10

Similar Documents

Publication Publication Date Title
Wang et al. A novel deep learning based fault diagnosis approach for chemical process with extended deep belief network
CN110414788B (en) Electric energy quality prediction method based on similar days and improved LSTM
Lu et al. Sales forecasting for computer wholesalers: A comparison of multivariate adaptive regression splines and artificial neural networks
Han et al. Food quality and safety risk assessment using a novel HMM method based on GRA
Guo et al. Output-relevant variational autoencoder for just-in-time soft sensor modeling with missing data
CN112597392B (en) Recommendation system based on dynamic attention and hierarchical reinforcement learning
CN111813084A (en) Mechanical equipment fault diagnosis method based on deep learning
CN108399434B (en) Analysis and prediction method of high-dimensional time series data based on feature extraction
Su et al. Prediction model of permeability index for blast furnace based on the improved multi-layer extreme learning machine and wavelet transform
CN110119540B (en) Multi-output gradient lifting tree modeling method for survival risk analysis
CN113780420B (en) GRU-GCN-based method for predicting concentration of dissolved gas in transformer oil
CN112989711B (en) Aureomycin fermentation process soft measurement modeling method based on semi-supervised ensemble learning
CN112560948B (en) Fundus image classification method and imaging method under data deviation
CN115115416A (en) Commodity sales forecasting method
CN112464567B (en) Intelligent data assimilation method based on variational and assimilative framework
CN113378383B (en) Food supply chain hazard prediction method and device
CN113657452A (en) Tobacco leaf quality grade classification prediction method based on principal component analysis and super learning
CN116303786B (en) Block chain financial big data management system based on multidimensional data fusion algorithm
CN116258087B (en) Matte grade soft measurement method and device, electronic equipment and storage medium
CN108009859A (en) Agricultural product price fluctuates method for early warning and equipment
CN116432856A (en) Pipeline dynamic early warning method and device based on CNN-GLSTM model
CN114358502A (en) Edible oil risk assessment method based on optimized ELM
CN110728024B (en) Vine copula-based soft measurement method and system
CN114463014A (en) SVM-Xgboost-based mobile payment risk early warning method
CN112651168B (en) Construction land area prediction method based on improved neural network algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant