CN111861552A - Agricultural product price prediction method based on SHD-ELM - Google Patents
Agricultural product price prediction method based on SHD-ELM Download PDFInfo
- Publication number
- CN111861552A CN111861552A CN202010674881.0A CN202010674881A CN111861552A CN 111861552 A CN111861552 A CN 111861552A CN 202010674881 A CN202010674881 A CN 202010674881A CN 111861552 A CN111861552 A CN 111861552A
- Authority
- CN
- China
- Prior art keywords
- agricultural product
- prediction
- decomposition
- product price
- price
- 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.)
- Withdrawn
Links
Images
Classifications
-
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0283—Price estimation or determination
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0283—Price estimation or determination
- G06Q30/0284—Time or distance, e.g. usage of parking meters or taximeters
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Development Economics (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an agricultural product price prediction method based on SHD-ELM, which comprises the following steps: firstly, collecting agricultural product price time sequence data; decomposing the original agricultural product price time sequence into a plurality of Intrinsic Mode Functions (IMF) and remainder terms by empirical mode decomposition; secondly, performing secondary mixed decomposition on the influence of the irregularity of the IMF1 component with the most severe fluctuation on the prediction, namely performing wavelet transform decomposition on the IMF1 to obtain an approximate sequence and a detail sequence; predicting all the components obtained after decomposition by using an extreme learning machine; finally, combining the prediction results of all the components to obtain a prediction value of the original agricultural product price time sequence; the method accurately predicts the price of the agricultural products, and the prediction error is extremely small; compared with prediction methods such as a BP neural network and the like, the prediction method combining empirical mode decomposition, wavelet transformation and extreme learning machines has better agricultural product price prediction performance and can be suitable for predicting the price fluctuation rule of agricultural products.
Description
Technical Field
The invention belongs to the technical field of agricultural product data processing, and particularly relates to an SHD-ELM-based agricultural product price prediction method.
Background
The price prediction of the agricultural products belongs to the category of time series prediction, and meanwhile, the agricultural products have the particularity of high corrosion susceptibility and need to meet the supply and demand balance, so that the price prediction of the agricultural products is different from the price prediction of common commodities. In reality, a plurality of external factors such as climate change, economic fluctuation, special holidays and the like all affect the price of the agricultural products, so that the price of the agricultural products shows high random fluctuation. This also makes high accuracy agricultural product price prediction quite challenging. By developing the analysis and prediction of the price of the agricultural products, prospective information is provided for farmers and production operators, production and sales ideas are adjusted in time, loss caused by abnormal fluctuation of the price of the agricultural products is prevented, the risk evading capability of the agricultural market is improved, decision reference is provided for the local government to make stable agricultural product market supply and demand and price-related policies, and meanwhile, reference is provided for consumers to select daily diet consumption.
Agricultural product planting is limited and influenced by climate, and most of agricultural product production has obvious cycle characteristics. The price of the agricultural products is mainly influenced by market supply and demand, and shows characteristics of seasonality, tendency, random fluctuation and the like, and the time sequence of the agricultural products shows typical nonlinear characteristics. Price prediction methods at home and abroad are mainly divided into 4 types: a measurement economy prediction method, a mathematical statistics prediction method, an intelligent analysis method and a combined model method. In order to disclose a causal relationship in an economic phenomenon, the most common method in agricultural product market price prediction is a regression analysis method, and a linear or nonlinear regression model is generally established according to a functional relationship between an independent variable and a dependent variable, and then the corresponding agricultural product price is predicted. However, when the price of an agricultural product is predicted, the factors influencing the price variation are more, and cannot be listed one by one during the prediction, and the individual influencing factors are not quantized, which brings great inconvenience to the price prediction work. Statistical prediction is an important function of statistics and an important method in the prediction field. In the market price prediction of agricultural products, the most widely applied method is the traditional time series prediction method which mainly comprises a seasonal index method, a moving average method, an index smoothing method and the like. The traditional time series prediction method has higher precision when predicting time series data with linear characteristics, but when the market gradually develops towards complication, nonlinearity and irregularity, the method brings great difficulty for model parameter adjustment when predicting prices, and reflects certain limitation. With the wide application of information technology and intelligent technology in various fields, the price prediction of agricultural products is also carried out meaningful practice by applying the new technologies. The intelligent analysis method applied to agricultural product market price prediction mainly comprises a neural network prediction method, a gray prediction method, a support vector machine and the like. The intelligent analysis method has strong nonlinear mapping capability and flexible network structure when predicting the time sequence, and simultaneously has the advantage of minimizing the sample prediction error. The combinatorial model method is a new prediction method formed by combining different methods. It has two basic forms, one is to combine the results of different prediction methods with equal weight or unequal weight. And secondly, before a prediction method is used for prediction, the combination is processed by an auxiliary method so that the prediction precision of the subsequent prediction method is better. The results of establishing the combined prediction model prove to be higher than the accuracy of a single prediction model, and then become the trend of the prediction field.
The extreme learning machine is a machine learning method with excellent performance proposed by Huang in 2006, is a single hidden layer feedforward neural network essentially, can greatly improve generalization capability and learning speed of the network compared with the traditional neural network, and has strong nonlinear fitting capability. Extreme learning machines are now widely used in stock prices, network traffic, wind speed prediction, and the like. However, the price sequence of the agricultural product is a special sequence signal with nonlinearity and non-stationarity, and although the extreme learning machine can well fit the nonlinear part of the price sequence, the non-stationarity part of the price of the agricultural product has a large influence on the prediction effect.
Empirical mode decomposition can decompose a complex non-stationary signal into different signals with local time-varying characteristics, so that the non-stationary of a sequence is effectively reduced.
The wavelet transformation inherits and develops the idea of Fourier transformation localization, can carry out multi-scale analysis on signals, can fully highlight the detail characteristics of the signals in the decomposition process, carry out time subdivision on the high-frequency part of the signals and carry out frequency subdivision on the low-frequency part of the signals, thereby achieving the effect of focusing the detail characteristics of the signals and realizing more detailed approximation on an analysis object.
Therefore, the invention provides an agricultural product price prediction method based on SHD-ELM.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides an agricultural product price combination prediction method with more accurate prediction precision, combines empirical mode decomposition, wavelet transformation decomposition and extreme learning machines to predict the price of an agricultural product, has smaller prediction error, and can be suitable for predicting the price fluctuation rule of the agricultural product.
In order to achieve the purpose, the invention provides the following technical scheme: an agricultural product price prediction method based on SHD-ELM comprises the following steps: firstly, collecting agricultural product price time sequence data; decomposing the original agricultural product price time sequence into a plurality of Intrinsic Mode Functions (IMF) and remainder terms by empirical mode decomposition; secondly, performing secondary mixed decomposition on the influence of the irregularity of the IMF1 component with the most severe fluctuation on the prediction, namely performing wavelet transform decomposition on the IMF1 to obtain an approximate sequence and a detail sequence; predicting all the components obtained after decomposition by using an extreme learning machine; and finally, combining the prediction results of all the components to obtain the prediction value of the original agricultural product price time sequence.
Further, the agricultural product price time sequence data are data at equal intervals.
Further, all the components obtained after the decomposition include other Intrinsic Mode Functions (IMFs) except the IMF1 component, remainder, approximate sequence obtained by the IMF1 decomposition, and detail sequence.
Further, the empirical mode decomposition comprises the following steps: (1) time sequence of taking original agricultural product price
All maxima and minima of; (2) fitting the upper envelope curves of the maximum value and the minimum value through cubic spline functions respectively
And a lower envelope
Calculating the mean of the upper and lower envelope
I.e. by
(ii) a (3) Computing
And
the difference of the difference is recorded as
，
=
-
If, if
Two conditions satisfying the eigenmode function are recorded
=
，
Is the first eigenmode function component; if it is not
Not an intrinsic mode function, will
Viewed as a new signal
Repeating the above steps until
Is an intrinsic mode function, and is recorded as
(ii) a (4) By mixing the remaining part
=
-
The steps are repeated as a new signal, and the rest intrinsic mode functions and a remainder can be extracted; raw data
Can be expressed as an eigenmode function andsum of the remainder, i.e.
=
(ii) a Thus, a time signal can be decomposed into 
The sum of the eigenmode function components, which contain components in different frequency bands, and a remainder.
Further, the intrinsic mode functions are conditioned by two conditions: (1) in the whole signal sequence, the difference between the number of extreme points and the number of zero-crossing points is not more than 1; (2) the average of the envelope of the local maximum value and the envelope of the local minimum value at an arbitrary time point is 0.
Further, the wavelet transform adopts Daubechies wavelet basis function, dbN for short, N is the order of the wavelet, the order is selected to be 5, and denoising processing is performed on the selected IMF1 component.
The invention has the following beneficial effects:
according to the method, different characteristic components in an original price sequence can be effectively separated by adopting an empirical mode decomposition method according to the specific fluctuation form of the price of the agricultural product, and the non-stationarity of the original sequence can be well reduced by utilizing secondary mixed decomposition, so that the prediction precision of the price of the agricultural product is improved; the invention establishes a combined prediction method based on empirical mode Decomposition (SHD) and quadratic Hybrid Decomposition (SHD), and the method has better prediction capability compared with BP neural network and single SHD method, and more accurate and reliable prediction compared with a prediction method without wavelet transformation; a new idea is provided for nonlinear and complex agricultural product price prediction; the short-term accurate prediction of the price of the agricultural products is simulated for the future price trend, and the method has important practical significance for agricultural product producers, operators, consumers and relevant government departments.
Drawings
FIG. 1 is a flow chart of the present invention.
Fig. 2 is a graph of potato price trends collected in an embodiment of the present invention.
FIG. 3 is a graph of potato price sequences and EMD components for an embodiment of the present invention.
Fig. 4 is a diagram of IMF1 and wavelet transform components in an embodiment of the present invention.
FIG. 5 is a diagram of the input and output data structures of an ELM in an embodiment of the present invention.
FIG. 6 is a comparison of the predicted effects of different models in an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1 to 6, the present embodiment provides a method for predicting prices of agricultural products based on SHD-ELM, including the following steps: firstly, collecting agricultural product price time sequence data; decomposing the original agricultural product price time sequence into a plurality of Intrinsic Mode Functions (IMF) and remainder terms by empirical mode decomposition; secondly, performing secondary mixed decomposition on the influence of the irregularity of the IMF1 component with the most severe fluctuation on the prediction, namely performing wavelet transform decomposition on the IMF1 to obtain an approximate sequence and a detail sequence; predicting all the components obtained after decomposition by using an extreme learning machine; and finally, combining the prediction results of all the components to obtain the prediction value of the original agricultural product price time sequence.
In this embodiment, the agricultural product price time-series data are data at equal intervals.
In this embodiment, all the components obtained after the decomposition include an Intrinsic Mode Function (IMF) other than the IMF1 component, a remainder, an approximate sequence obtained by the decomposition of the IMF1, and a detail sequence.
In the present embodimentThe steps of the empirical mode decomposition are as follows: (1) time sequence of taking original agricultural product price
All maxima and minima of; (2) fitting the upper envelope curves of the maximum value and the minimum value through cubic spline functions respectively
And a lower envelope
Calculating the mean of the upper and lower envelope
I.e. by
(ii) a (3) Computing
And
the difference of the difference is recorded as
，
=
-
If, if
Two conditions satisfying the eigenmode function are recorded
=
，
Is the first eigenmode function component; if it is not
Not an intrinsic mode function, will
Viewed as a new signal
Repeating the above steps until
Is an intrinsic mode function, and is recorded as
(ii) a (4) By mixing the remaining part
=
-
The steps are repeated as a new signal, and the rest intrinsic mode functions and a remainder can be extracted; raw data
Can be expressed as the sum of the eigenmode function and the remainder, i.e.
=
(ii) a Thus, a time signal can be decomposed into 
The sum of the eigenmode function components, which contain components in different frequency bands, and a remainder.
In this embodiment, the intrinsic mode function has two conditions: (1) in the whole signal sequence, the difference between the number of extreme points and the number of zero-crossing points is not more than 1; (2) the average of the envelope of the local maximum value and the envelope of the local minimum value at an arbitrary time point is 0. Empirical Mode Decomposition (EMD) is a signal decomposition method suitable for a nonlinear and non-stationary time sequence in Hilber-Huang transformation; the method fundamentally gets rid of the constraint of the Fourier transform theory, decomposes the signal according to the time scale characteristics of the signal, does not need to set any basis function in advance, and is essentially different from the Fourier decomposition and wavelet decomposition methods. Empirical mode decomposition decomposes the original sequence into eigenmode functions with different scales, stationarity and periodic fluctuation characteristics and a residual component representing the general trend of the original signal.
In this embodiment, the wavelet transform adopts Daubechies wavelet basis function, dbN for short, N is the order of the wavelet, the order is selected to be 5, and denoising processing is performed on the selected IMF1 component. According to the distribution characteristics of the price time sequence of the agricultural product, a wavelet transformation method is adopted for denoising, the prediction precision of time sequence signals is improved, mainly because the wavelet transformation can simultaneously observe frequency and a time axis, fine or rough parts of the signals can be separated, and the selection of wavelet bases has the advantages of diversity and the like.
The extreme learning machine is used as a novel single hidden layer neural network, and can overcome the defects that the traditional feedforward neural network is slow in training speed, network parameters are updated repeatedly, and local optimization is easy to fall into. The network structure of the extreme learning machine is basically the same as that of a single hidden layer feedforward neural network, and the extreme learning machine comprises an input layer, an output layer and a hidden layer. The solution obtained by the extreme learning machine is the only optimal solution, the network generalization capability is ensured, the learning speed is high, and the like, and on the premise of ensuring higher precision and prediction effect, the excellent training speed provides convenience for combining the extreme learning machine with other methods.
The invention has the advantages of simplicity and rapidness in application. The combined prediction model based on empirical mode decomposition, quadratic hybrid decomposition and extreme learning machine can automatically decompose each intrinsic mode function and a remainder according to the time scale of the sequence to be predicted by fully utilizing the empirical mode decomposition method, the number of the intrinsic mode functions is specifically generated according to the characteristics of the original sequence, and the number of the intrinsic mode functions possibly generated by different original sequences is different. The extreme learning machine solves the problem that a large number of network parameters need to be set in the traditional neural network, and the sequence can be learned and predicted only by setting the number of the neurons of the hidden layer in the application process.
Referring to fig. 1, firstly, according to an original sequence of the price of an agricultural product, decomposition is performed according to the time scale of the original sequence by using an empirical mode decomposition method to obtain a plurality of intrinsic mode function components and a remainder component, and the frequency component of each intrinsic mode function component is closely related to the original sequence and changes along with the change of the intrinsic mode function component, so that the method has self-adaptability. The rest items are decomposed to represent the fluctuation trend of the original agricultural product price sequence. The IMF1 is subjected to wavelet transformation to decompose an approximate sequence and a detail sequence, all decomposed components are respectively predicted by using an extreme learning machine method, and each component is set with different parameter values through the extreme learning machine according to the characteristics of the component to obtain different prediction results. And finally, combining and reconstructing the prediction results of all the components to obtain a final agricultural product price prediction value.
The experimental data of the embodiment are from an agricultural product information monitoring system in Henan province, monthly average price data of certain farmer market potatoes are selected from the system, the availability and the continuity of samples are considered, a total of 72 data samples with a data period of 2014 to 2019 and 12 months are finally selected, the data are price time sequence data with equal intervals, and the sequence is unstable and fluctuates irregularly along with time as can be seen from FIG. 2. In the experimental study, 60 samples from 1 month 2014 to 12 months 2018 are used as a training data set, and refer to fig. 2; a 12-month sample of 2019 was used as a test data set for application verification of the predictive model.
Evaluation criteria: to evaluate the prediction effectiveness of the combined prediction model, Mean Absolute Error (MAE), mean percent error (MAPE), and Root Mean Square Error (RMSE) indicators were selected to measure the accuracy of the prediction. The calculation formula of each prediction index is as follows: (1)
；（2）
；（3）
(ii) a In the formula, the first step is that,
for the number of price samples to be taken,
is as follows
Monthly average price of months.
The price trend graph of the potato shows that the sequence has unstable and nonlinear characteristics, so that an empirical mode decomposition method can be used for decomposing the price time sequence, and MatlabR2014b is used for calling an empirical mode decomposition tool box writing program to decompose the price time sequence to obtain 4 intrinsic mode function components and 1 linear trend margin. The decomposition result is shown in fig. 3, the first graph in the graph is an original sequence graph of the potato, the middle 4 are intrinsic mode function component graphs, and the last graph is a margin graph, so that the graph can find that the fluctuation of the IMF1 component in the intrinsic mode function component is severe, the fluctuation of the IMF2 component is smooth and stable as a whole, the IMF3 and the IMF4 component change gradually and slowly and have strong regularity, and the margin is consistent with the whole trend of the original price sequence, so that the linear trend of the price sequence is reflected.
And a secondary decomposition method is proposed for solving the problem that the IMF1 component still fluctuates severely after decomposition, namely, IMF1 is decomposed again by using wavelet transformation. According to the specific fluctuation characteristics of the IMF1 component, a Daubechies wavelet basis function, dbN for short, N is the order of the wavelet, the order is selected to be 5, and the selected IMF1 component is subjected to denoising treatment. When the decomposition scale is too small, the salient of detailed characteristic signals of the price sequence of the agricultural product under the condition of multiple factors is influenced, the prediction precision is not improved favorably, and when the decomposition scale is too large, the practical process operation is aggravated. In the embodiment, the decomposition scale of the wavelet is selected to be 2, so that the problem caused by too large or too small decomposition scale can be avoided. After the decomposition scale is determined, 2-scale decomposition of the IMF1 component is carried out on db5 to obtain a trend low-frequency part a2 and a detail high-frequency part d1 and d2, and a more detailed perspective is carried out on the frequency and the amplitude of fluctuation of the IMF1, which is shown in FIG. 4.
And respectively predicting the decomposed components by using an extreme learning machine, in order to realize higher prediction precision, firstly carrying out normalization processing on price data in the prediction process, wherein the decomposed continuous prices from 2014 to 2018 are used as a training set of the model, and the decomposed continuous prices from 2019 in the month are used as a testing set. When the neural network is set, seven data including the current price of six months and the price of the same month in the previous year are used as input layer data to predict the price in one step in advance, and the input and output structures are shown in fig. 5. The number of the neurons of the hidden layer is selected from different values through practice according to the fluctuation characteristics of different components. And performing inverse normalization processing on the output data, reducing the output data into price data, and finally performing accumulation combination on the predicted values of all the components to obtain the final predicted value. The modeling experiment of the extreme learning machine is implemented by programming MatlabR2104 b.
TABLE-COMPARISON OF PREDICTION ERRORS OF DIFFERENT MODELS
| Evaluation index | SHD-ELM | EMD-ELM | ELM | BP |
| MAE (Yuan/kg) | 0.064 | 0.093 | 0.182 | 0.214 |
| MAPE（%） | 2.702 | 4.265 | 7.62 | 9.68 |
| RMSE (Yuan/kg) | 0.084 | 0.148 | 0.233 | 0.236 |
In order to compare the prediction performance, the same data set is modeled and predicted by adopting a BP neural network and a single extreme learning machine respectively, wherein the input data volume of the BP neural network and the input data volume of the extreme learning machine are consistent with the prediction setting of a combined model, and seven data including the current price of six months and the previous month price of the year are taken as the input data set to be predicted in advance by one step. As can be seen from fig. 6 and table one, the prediction accuracy of the SHD-ELM model is greatly improved from the prediction result of the single extreme learning machine and the BP neural network, and is superior to the combined model result of the empirical mode decomposition and the extreme learning machine, and since the original agricultural product price sequence is decomposed by the empirical mode decomposition method and decomposed again by wavelet transformation for the IMF1, the complexity of the subsequent prediction process is reduced by all the decomposed components. Specifically, from the precision evaluation index, the average absolute error of the prediction model of the quadratic hybrid decomposition and the extreme learning machine is 0.064 yuan/kg, the average absolute error of the combined prediction model of the empirical mode decomposition and the extreme learning machine is 0.093 yuan/kg, the prediction results of the BP neural network and the single extreme learning machine are 0.214 yuan/kg and 0.182 yuan/kg respectively, and the prediction results of the BP neural network and the extreme learning machine do not have the combined prediction model of the invention to perform well compared with the average percentage error and the root mean square error. And the prediction based on the quadratic hybrid decomposition and the extreme learning machine is better than the combined prediction of the empirical mode decomposition and the extreme learning. Therefore, in the combined prediction model, the quadratic hybrid decomposition can well reduce the non-stationarity of the original sequence, and the prediction precision of the price of agricultural products is improved.
In conclusion, according to the method, different characteristic components in the original price sequence can be effectively separated by adopting an empirical mode decomposition method according to the specific fluctuation form of the price of the agricultural product, and the IMF1 components are subjected to wavelet transformation for secondary decomposition, so that the non-stationarity of each component is effectively reduced. The invention establishes a combined prediction model based on quadratic hybrid decomposition and extreme learning machine, and experimental results show that the prediction method has better prediction capability than BP neural network, single extreme learning machine method and the combined model of empirical mode decomposition and extreme learning machine, and provides a new thought for nonlinear complex agricultural product price prediction; the short-term accurate prediction of the price of the agricultural product is simulated for the future price trend, which has important practical significance for agricultural product producers, operators, consumers and relevant government departments.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (6)
1. An agricultural product price prediction method based on SHD-ELM is characterized by comprising the following steps: firstly, collecting agricultural product price time sequence data; decomposing the original agricultural product price time sequence into a plurality of Intrinsic Mode Functions (IMF) and remainder terms by empirical mode decomposition; secondly, performing secondary mixed decomposition on the influence of the irregularity of the IMF1 component with the most severe fluctuation on the prediction, namely performing wavelet transform decomposition on the IMF1 to obtain an approximate sequence and a detail sequence; predicting all the components obtained after decomposition by using an extreme learning machine; and finally, combining the prediction results of all the components to obtain the prediction value of the original agricultural product price time sequence.
2. The SHD-ELM-based agricultural product price forecasting method of claim 1, wherein: the agricultural product price time sequence data are data at equal intervals.
3. The SHD-ELM-based agricultural product price forecasting method of claim 2, wherein: all the components obtained after the decomposition comprise other Intrinsic Mode Functions (IMF) except the IMF1 component, remainder, approximate sequences and detail sequences obtained by IMF1 decomposition.
4. The SHD-ELM-based agricultural product price forecasting method of claim 3, wherein: the empirical mode decomposition comprises the following steps: (1) time sequence of taking original agricultural product price 
All maxima and minima of; (2) fitting the upper envelope curves of the maximum value and the minimum value through cubic spline functions respectively
And a lower envelope
Calculating the mean of the upper and lower envelope
I.e. by
(ii) a (3) Computing
And
the difference of the difference is recorded as
，
=
-
If, if
Two conditions satisfying the eigenmode function are recorded
=
，
Is the first eigenmode function component; if it is not
Not an intrinsic mode function, will
Viewed as a new signal
Repeating the above steps until
Is an intrinsic mode function, and is recorded as
(ii) a (4) By mixing the remaining part
=
-
The steps are repeated as a new signal, and the rest intrinsic mode functions and a remainder can be extracted; raw data
Can be expressed as the sum of the eigenmode function and the remainder, i.e.
=
(ii) a Thus, a time signal can be decomposed into
The sum of the eigenmode function components, which contain components in different frequency bands, and a remainder.
5. The SHD-ELM-based agricultural product price forecasting method of claim 4, wherein: two conditions of the eigenmode function: (1) in the whole signal sequence, the difference between the number of extreme points and the number of zero-crossing points is not more than 1; (2) the average of the envelope of the local maximum value and the envelope of the local minimum value at an arbitrary time point is 0.
6. The SHD-ELM-based agricultural product price forecasting method of claim 5, wherein: the wavelet transform adopts a Daubechies wavelet basis function, dbN for short, N is the order of the wavelet, the order is selected to be 5, and denoising processing is carried out on the selected IMF1 component.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010674881.0A CN111861552A (en) | 2020-07-14 | 2020-07-14 | Agricultural product price prediction method based on SHD-ELM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010674881.0A CN111861552A (en) | 2020-07-14 | 2020-07-14 | Agricultural product price prediction method based on SHD-ELM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111861552A true CN111861552A (en) | 2020-10-30 |
Family
ID=72984065
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010674881.0A Withdrawn CN111861552A (en) | 2020-07-14 | 2020-07-14 | Agricultural product price prediction method based on SHD-ELM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111861552A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508088A (en) * | 2020-12-03 | 2021-03-16 | 重庆邮智机器人研究院有限公司 | DEDBN-ELM-based electroencephalogram emotion recognition method |
| CN113378387A (en) * | 2021-06-10 | 2021-09-10 | 大连海事大学 | Ship motion forecasting method based on improved EMD-AR model |
| CN113409072A (en) * | 2021-05-31 | 2021-09-17 | 河北科技师范学院 | Empirical mode decomposition and distributed GRU neural network and price prediction method |
| CN115358475A (en) * | 2022-08-29 | 2022-11-18 | 河南农业大学 | Disaster prediction method and system based on support vector machine and gray BP neural network |
-
2020
- 2020-07-14 CN CN202010674881.0A patent/CN111861552A/en not_active Withdrawn
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508088A (en) * | 2020-12-03 | 2021-03-16 | 重庆邮智机器人研究院有限公司 | DEDBN-ELM-based electroencephalogram emotion recognition method |
| CN113409072A (en) * | 2021-05-31 | 2021-09-17 | 河北科技师范学院 | Empirical mode decomposition and distributed GRU neural network and price prediction method |
| CN113378387A (en) * | 2021-06-10 | 2021-09-10 | 大连海事大学 | Ship motion forecasting method based on improved EMD-AR model |
| CN113378387B (en) * | 2021-06-10 | 2022-11-01 | 大连海事大学 | Ship motion forecasting method based on improved EMD-AR model |
| CN115358475A (en) * | 2022-08-29 | 2022-11-18 | 河南农业大学 | Disaster prediction method and system based on support vector machine and gray BP neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111861552A (en) | Agricultural product price prediction method based on SHD-ELM | |
| CN110135612B (en) | Method for monitoring capacity of material supply Shang Sheng and early warning abnormality based on electricity consumption analysis | |
| CN112488395B (en) | Method and system for predicting line loss of power distribution network | |
| Deng et al. | Inter-company comparison using modified TOPSIS with objective weights | |
| CN108710905B (en) | Spare part quantity prediction method and system based on multi-model combination | |
| US20210103858A1 (en) | Method and system for model auto-selection using an ensemble of machine learning models | |
| CN111489259A (en) | Stock market risk prediction intelligent implementation method based on deep learning | |
| CN110533249B (en) | Metallurgical enterprise energy consumption prediction method based on integrated long-term and short-term memory network | |
| CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
| CN111563628A (en) | Real estate customer transaction time prediction method, device and storage medium | |
| CN116826745B (en) | Layered and partitioned short-term load prediction method and system in power system background | |
| Ilich et al. | Modeling of employment structural transformations | |
| CN112598181A (en) | Load prediction method, device, equipment and storage medium | |
| Hasheminejad et al. | Developing cluster-based adaptive network fuzzy inference system tuned by particle swarm optimization to forecast annual automotive sales: a case study in Iran market | |
| CN111861553A (en) | Agricultural product price prediction method based on EMD-ELM | |
| Gurbaxani | The demand for information technology capital An empirical analysis | |
| CN115564265A (en) | Power grid enterprise digital transformation evaluation method based on evaluation model | |
| Sun et al. | Short-term stock price forecasting based on an SVD-LSTM model | |
| CN115115414A (en) | Second-hand car valuation method based on machine learning | |
| Fan et al. | Multi-model fusion based on stacking: a predictive model for the price trend of natural rubber | |
| Ergemen | Parametric estimation of long memory in factor models | |
| Amiri et al. | Differentiation in Differentiation Method and Neural Network to Investigate Different Criteria of Economic Productivity | |
| Lukić | Merenje i analiza dinamike profitabilnosti bankarskog sektora u Srbiji na bazi FLMAW-MARCOS metoda | |
| Supriyanto | Comparison of Grid Search and Evolutionary Parameter Optimization with Neural Networks on JCI Stock Price Movements during the Covid 19 | |
| Wu et al. | Research on Economic Forecast of High-Tech Park Based on Combination Model: Taking the Zhongguancun Demonstration Zone as an Example |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20201030 |