Method, device and equipment for interval prediction of gasoline business data and storage medium

Abstract

The invention relates to the field of gasoline data prediction, and particularly provides a method, device and equipment for interval prediction of gasoline business data and a storage medium, the method comprises the following steps: decomposing historical business data of to-be-predicted gasoline through a binary empirical mode decomposition method to obtain a plurality of business data fluctuation curves of different frequencies; historical business data in the business data fluctuation curves of different frequencies are input into a plurality of preset business data prediction models, a plurality of initial business data prediction results and residual results are obtained, and the business data prediction models comprise a threshold autoregression interval model or a target business data prediction model; and determining a final business data prediction result of the to-be-predicted gasoline according to the plurality of initial business data prediction results and the residual result. Through the method, the effect of accurately predicting the gasoline service data can be achieved.

Description

Technical Field

[0001] The present application relates to the field of gasoline data prediction, and in particular, to a method, device, equipment and storage medium for interval prediction of gasoline business data. Background Art

[0002] In recent years, with the increasing global awareness of environmental protection and the advancement of low-carbon economy, the iterative upgrade of carbon emission reduction technology has been promoted, which has affected the fluctuation of gasoline-related business data (for example, gasoline price trend forecast, demand forecast, sales data analysis and production data). Gasoline business data presents nonlinear and irregular complex fluctuations, making prediction more challenging and complex. Therefore, under the global low-carbon development trend, predicting business data as an effective tool for managing energy allocation and reducing carbon emissions has become particularly important.

[0003] Traditional point-based forecasting methods, such as econometric forecasting methods, artificial intelligence-based forecasting methods, and hybrid forecasting methods, are prone to losing important fluctuation information and are easily affected by noise and small shocks, which affects the forecast accuracy.

[0004] Therefore, how to accurately predict gasoline business data is a technical problem that needs to be solved. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide a method for interval prediction of gasoline business data. The technical solution of the embodiments of the present application can achieve the effect of accurately predicting gasoline business data.

[0006] In a first aspect, an embodiment of the present application provides a method for interval prediction of gasoline business data, comprising: decomposing the historical business data of the gasoline to be predicted by a binary empirical mode decomposition method to obtain a plurality of business data fluctuation curves of different frequencies; inputting the historical business data in the plurality of business data fluctuation curves of different frequencies into a plurality of preset business data prediction models to obtain a plurality of initial business data prediction results and residual results, wherein the business data prediction model comprises: a threshold autoregressive interval model or a target business data prediction model; determining the final business data prediction result of the gasoline to be predicted through the plurality of initial business data prediction results and residual results.

[0007] In the above embodiment of the present application, the captured gasoline business data is decomposed by the binary empirical mode decomposition method, the interval value data is divided into different frequency signals for analysis, and a suitable model is selected for prediction according to the frequency characteristics of different frequency intervals, thereby avoiding the loss of volatility information due to the complexity of the overall time series, improving the prediction accuracy and capturing more variability and uncertainty information, greatly improving the out-of-sample prediction of gasoline business data, and achieving the effect of accurately predicting gasoline business data.

[0008] In some embodiments, before decomposing the historical business data of the gasoline to be predicted by a binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies, it also includes: collecting all historical business data of various types of gasoline, and decomposing all historical business data by a binary empirical mode decomposition method to obtain a set of all historical business data; classifying the historical business data in the set of all historical business data according to a preset frequency to obtain multiple classification sets, wherein the classification set includes a high-frequency business data set, a medium-frequency business data set or a low-frequency business data set; training a target business data prediction model by using the high-frequency business data set and the medium-frequency business data set in multiple classification sets; and training a threshold autoregressive interval model by using the low-frequency business data set in multiple classification sets.

[0009] In the above embodiments of the present application, various types of gasoline historical business data are used as training samples. By decomposing and classifying the samples, classified set samples of different frequencies can be obtained. The corresponding prediction models are trained respectively by the classified set samples of different frequencies. The prediction models obtained by training can accurately predict the gasoline business data.

[0010] In some embodiments, a target business data prediction model is trained using a high-frequency business data set and a medium-frequency business data set in multiple classification sets, including: adding exogenous variables to the high-frequency business data set and the medium-frequency business data set to obtain a modified high-frequency business data set and a modified medium-frequency business data set, wherein the exogenous variables represent standard carbon emission data set set in the carbon reduction technology; inputting the modified high-frequency business data set and the modified medium-frequency business data set into a basic data prediction model to obtain prediction results; and adjusting the parameters of the basic data prediction model based on the prediction results to obtain a target business data prediction model.

[0011] In the above embodiments of the present application, the target business data prediction model can be trained through high and medium frequency gasoline business data samples to accurately predict high and medium frequency gasoline business data.

[0012] In some embodiments, a threshold autoregressive interval model is trained using a low-frequency business data set in multiple classification sets, including: inputting low-frequency business data in the low-frequency business data set into a basic threshold autoregressive interval model to obtain an output residual result; adjusting the parameters of the basic threshold autoregressive interval model using the output residual result and the standard residual result to obtain the threshold autoregressive interval model.

[0013] In the above embodiment of the present application, a threshold autoregressive interval model is trained by using low-frequency gasoline business data samples to accurately predict the residual of low-frequency gasoline business data.

[0014] In some embodiments, historical business data in multiple business data fluctuation curves of different frequencies are input into multiple preset business data prediction models, including: inputting high-frequency business data and medium-frequency business data in multiple business data fluctuation curves of different frequencies into a target business data prediction model to obtain multiple initial business data prediction results; inputting low-frequency business data in multiple business data fluctuation curves of different frequencies into a threshold autoregressive interval model to obtain residual results.

[0015] In the above embodiments of the present application, the target business data prediction model and the threshold autoregressive interval model can respectively accurately predict the business data prediction results of the gasoline to be predicted and the residual results of the gasoline business data.

[0016] In some embodiments, the final business data prediction result of the gasoline to be predicted is determined through multiple initial business data prediction results and residual results, including: adding multiple initial business data prediction results and residual results to obtain the final business data prediction result, wherein the business data prediction result includes: gasoline price trend forecast, demand forecast, sales data analysis or production data.

[0017] In the above embodiment of the present application, by superimposing the prediction results of the target business data prediction model and the threshold autoregressive interval model respectively, the final prediction result of the gasoline business data can be accurately obtained.

[0018] In a second aspect, an embodiment of the present application provides a device for interval prediction of gasoline business data, comprising: A decomposition module is used to decompose the historical business data of gasoline to be predicted by using a binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies; A prediction module is used to input historical business data in multiple business data fluctuation curves of different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model; The determination module is used to determine the final business data prediction result of the gasoline to be predicted through multiple initial business data prediction results and residual results.

[0019] Optionally, the device further comprises: A training module, used for the decomposition module to decompose the historical business data of the gasoline to be predicted by the binary empirical mode decomposition method to obtain a plurality of business data fluctuation curves of different frequencies, collect all historical business data of various types of gasoline, and decompose all historical business data by the binary empirical mode decomposition method to obtain a set of all historical business data; Classifying the historical business data in all historical business data sets according to preset frequencies to obtain multiple classification sets, wherein the classification sets include high-frequency business data sets, medium-frequency business data sets, or low-frequency business data sets; Training a target business data prediction model by using a high-frequency business data set and a medium-frequency business data set in a plurality of classification sets; The threshold autoregressive interval model is trained by using low-frequency business data sets in multiple classification sets.

[0020] Optionally, the training module is specifically used to: Adding exogenous variables to the high-frequency business data set and the medium-frequency business data set to obtain a modified high-frequency business data set and a modified medium-frequency business data set, wherein the exogenous variables represent standard carbon emission data set in the carbon emission reduction technology; Inputting the modified high-frequency business data set and the modified medium-frequency business data set into the basic data prediction model to obtain a prediction result; Based on the prediction results, the parameters of the basic data prediction model are adjusted to obtain the target business data prediction model.

[0021] Optionally, the training module is specifically used to: Adding multiple multilayer perceptron networks and estimators to the basic threshold autoregressive interval model to obtain a modified basic threshold autoregressive interval model; The modified basic threshold autoregressive interval model is trained by a low-frequency business data set to obtain a threshold autoregressive interval model.

[0022] Optionally, the decomposition module is specifically used to: The historical business data is filtered through preset time-frequency transformation parameters and frequency thresholds, and the filtered historical business data is decomposed to obtain multiple business data fluctuation curves of different frequencies.

[0023] Optionally, the prediction module is specifically used for: Inputting high-frequency business data and medium-frequency business data in a plurality of business data fluctuation curves of different frequencies into a target business data prediction model to obtain a plurality of initial business data prediction results; The low-frequency business data in a plurality of business data fluctuation curves with different frequencies are input into the threshold autoregressive interval model to obtain the residual result.

[0024] Optionally, determine whether the module is specifically used for: Multiple initial business data prediction results and residual results are added together to obtain a final business data prediction result, wherein the business data prediction result includes: gasoline price trend prediction, demand forecast, sales data analysis or production data.

[0025] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method provided in the first aspect are performed.

[0026] In a fourth aspect, an embodiment of the present application provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the method provided in the first aspect are performed.

[0027] Other features and advantages of the present application will be described in the subsequent description, and in part will become apparent from the description, or will be understood by implementing the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 A flowchart of a method for interval prediction of gasoline business data provided in an embodiment of the present application; Figure 2 A flowchart of an implementation method for interval prediction of gasoline prices provided in an embodiment of the present application; Figure 3 A schematic block diagram of a device for interval prediction of gasoline business data provided by an embodiment of the present application; Figure 4 A schematic diagram of the structure of a device for interval prediction of gasoline business data provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0031] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0032] First, some terms involved in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.

[0033] Bidimensional Empirical Mode Decomposition (BEMD) is a signal processing technique used to analyze two-dimensional data. BEMD decomposes two-dimensional data into a series of intrinsic mode functions (IMFs), which can capture the local characteristics and dynamic changes of the data.

[0034] The threshold autoregressive model, also known as the threshold model, is abbreviated as the TAR model. It is a nonlinear model used to solve a class of nonlinear problems. The threshold autoregressive interval model (TARI) is applied to low-frequency components through the method of this application. The TARI model is used to capture the nonlinear characteristics of interval-valued time series and is an interval extension of the classic TAR model.

[0035] Data residuals are the difference between observed values ​​and estimated values, and are used to measure the difference between actual data and model predicted data. In mathematical statistics and data analysis, residuals are an important concept that reflects the accuracy of the model in fitting the data. Specifically, residuals are particularly important in regression analysis because they can help us understand how well the model fits the data and what outliers may exist.

[0036] Friedman test, also known as Friedman two-way rank variance analysis, is a statistical test for the homogeneity of multiple (related) samples. The method was proposed by Friedman (M. Friedman) in 1973. Friedman test prerequisites: 1. Sequential data; 2. Three or more groups; 3. Related groups; 4. Randomly draw samples from the collocated values.

[0037] TFIDE is a comprehensive forecasting method that combines interval decomposition analysis and integrated forecasting technology to capture more information and improve out-of-sample forecasting. The advantages of this method are: First, the interval value data is divided into low, medium and high frequency signals for analysis, and the appropriate model is selected for forecasting according to the frequency characteristics of different frequency intervals. For economic and market phenomena that show nonlinear characteristics, nonlinear models (such as interval multilayer perceptron network (IMLP): Interval multilayer perceptron network) can be selected on the corresponding sub-intervals for forecasting to better adapt to nonlinear changes. Second, it avoids the loss of volatility information due to the complexity of the overall time series, improves forecasting accuracy, and captures more variability and uncertainty information, greatly improving the out-of-sample forecasting of gasoline price intervals. Third, it reduces the impact of noise and small shocks. By selecting appropriate time-frequency transformation parameters and thresholds, time-frequency analysis technology can treat noise and small shocks as high-frequency components and filter them out of the original time series. Then, forecasts are made on each sub-interval based on the data after noise filtering, thereby reducing the impact of noise and small shocks on forecast accuracy. In addition, due to the integration of the forecast results of multiple sub-intervals, the TFIDE method can further smooth the impact of these noises and small shocks and improve the stability of the forecast. Adapt to nonlinear changes. Fourth, adding two influencing factors, global carbon concerns and carbon reduction technology, to the original forecasting framework greatly improves the forecasting effect. Carbon reduction technology can help IMLP predict medium-frequency IMFs. At the same time, carbon concerns have a nonlinear impact on the long-term gasoline price range, which can reveal the close connection between the gasoline market and the low-carbon trend.

[0038] The DM test usually refers to the comparison of the prediction results of two models, and the performance of the model is evaluated by calculating the difference (d) of the prediction deviations of the two models. Specifically, the DM test uses different calculation formulas, such as mean square error (MSE), mean absolute error (MAD), mean percentage error (MAPE), etc., to calculate the prediction deviations of model 1 and model 2, and evaluates the quality of the model by comparing these deviations.

[0039] The present application is applied to the scenario of gasoline data prediction. The specific scenario is to decompose the historical business data of gasoline through the BEMD method to obtain historical gasoline business data of different frequencies, and use different models to predict the gasoline business data through the gasoline business data of different frequencies.

[0040] In recent years, with the increasing global awareness of environmental protection and the advancement of low-carbon economy, the iterative upgrade of carbon emission reduction technology has been promoted, which has affected the fluctuation of gasoline-related business data (for example, gasoline price trend forecast, demand forecast, sales data analysis and production data). Gasoline business data presents nonlinear and irregular complex fluctuations, making prediction more challenging and complex. Therefore, under the global low-carbon development trend, predicting business data as an effective tool for managing energy allocation and reducing carbon emissions has become particularly important. Traditional point-based forecasting methods, such as econometric forecasting methods, artificial intelligence-based forecasting methods and hybrid forecasting methods. Point forecasting methods are prone to lose important fluctuation information and are easily affected by noise and small shocks, affecting the prediction accuracy.

[0041] To this end, the present application decomposes the historical business data of gasoline to be predicted by a binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies; inputs the historical business data in the multiple business data fluctuation curves of different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model; through multiple initial business data prediction results and residual results, the final business data prediction results of gasoline to be predicted are determined. The captured gasoline business data is decomposed by a binary empirical mode decomposition method, and the interval value data is divided into signals of different frequencies for analysis. According to the frequency characteristics of different frequency intervals, a suitable model is selected for prediction, thereby avoiding the loss of volatility information due to the complexity of the overall time series, improving the prediction accuracy and capturing more variability and uncertainty information, greatly improving the out-of-sample prediction of gasoline business data, and achieving the effect of accurately predicting gasoline business data.

[0042] In the embodiment of the present application, the execution entity may be an interval prediction gasoline business data device in an interval prediction gasoline business data system. In actual applications, the interval prediction gasoline business data device may be electronic devices such as terminal devices and servers, which are not limited here.

[0043] Combine the following Figure 1 The method for interval prediction of gasoline business data in an embodiment of the present application is described in detail.

[0044] Please see Figure 1 , Figure 1A flowchart of a method for interval prediction of gasoline business data provided in an embodiment of the present application, such as Figure 1 The method for interval forecasting gasoline business data shown includes: Step 110: Decompose the historical business data of gasoline to be predicted by using a binary empirical mode decomposition method to obtain a plurality of business data fluctuation curves of different frequencies.

[0045] The historical business data includes: historical gasoline prices, historical gasoline demand, historical gasoline sales data analysis or historical gasoline production data.

[0046] In some embodiments of the present application, before decomposing the historical business data of the gasoline to be predicted by the binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies, it also includes: collecting all historical business data of various types of gasoline, and decomposing all historical business data by the binary empirical mode decomposition method to obtain a set of all historical business data; classifying the historical business data in the set of all historical business data according to preset frequencies to obtain multiple classification sets, wherein the classification sets include high-frequency business data sets, medium-frequency business data sets or low-frequency business data sets; training the target business data prediction model by the high-frequency business data sets and medium-frequency business data sets in the multiple classification sets; training the threshold autoregressive interval model by the low-frequency business data sets in the multiple classification sets.

[0047] In the above process, the present application uses various types of gasoline historical business data as training samples, and decomposes and classifies the samples to obtain classified set samples of different frequencies. The corresponding prediction models are trained by the classified set samples of different frequencies, and the prediction models obtained by training can accurately predict the gasoline business data.

[0048] Among them, the preset frequency can be set according to demand, for example, high-frequency words, medium frequency and low frequency, etc. All historical business data can collect historical business data of various types of gasoline, and various types of gasoline can be regular gasoline, unleaded gasoline, premium gasoline, oil-gas blend, medium volatility gasoline, mixed gasoline, diesel and natural gas, etc. When collecting data, 357 weeks of data points can be collected, and the data of the first 337 weeks are used to train the prediction model, and 10 weeks, 20 weeks and 30 weeks are used as model test data. When predicting gasoline price fluctuations, gasoline prices can be collected to construct interval weekly price data sets. That is, the maximum and minimum values ​​of the weekly interval price are found from the weekly observable interval to determine the range of price fluctuations. This data set contains all price fluctuation information in the time series. As model training data, it can effectively prevent information loss and improve the accuracy of prediction.

[0049] Optionally, when training a threshold autoregressive interval model or a target business data prediction model for predicting gasoline price fluctuations, the accuracy of the model can be improved by constructing a complex valued signal of the model. The specific construction method is as follows: Given a gas price sequence with interval values , Y represents the price, L, U represent the sequence number, construct a complex value signal ,in The specific operations are as follows: The interval-valued variable X represents the Define the corresponding relationship to R, so that for each t, , , = {[a, b]: a, b∈ Indicated by In this application, the gasoline price ,in and ( < ) are the lower and upper bounds of the weekly gasoline price respectively. And t, where t=1,2,3,……,n, represents time. Therefore, the interval value time series of this application is as follows:

[0050] Interval-valued time series data uses midpoint and half distance This can be expressed as:

[0051] This application uses the expression of upper and lower bounds as symbols. Define the operation rules between:

[0052]

[0053]

[0054] The minimum distance description is measured, where It's distance. Indicates price rules.

[0055]

[0056] The unit space is = . The symmetric positive definite weighting function Application To ensure It is a measure of the extended interval to construct a complex-valued signal.

[0057] In some embodiments of the present application, a target business data prediction model is trained by using a high-frequency business data set and a medium-frequency business data set in multiple classification sets, including: adding exogenous variables to the high-frequency business data set and the medium-frequency business data set to obtain a modified high-frequency business data set and a modified medium-frequency business data set, wherein the exogenous variables represent standard carbon emission data set set in the carbon reduction technology; inputting the modified high-frequency business data set and the modified medium-frequency business data set into a basic data prediction model to obtain prediction results; and adjusting the parameters of the basic data prediction model according to the prediction results to obtain a target business data prediction model.

[0058] In the above process, the present application can train the target business data prediction model through high and medium frequency gasoline business data samples to accurately predict the high and medium frequency gasoline business data.

[0059] In some embodiments of the present application, a threshold autoregressive interval model is trained using a low-frequency business data set in multiple classification sets, including: inputting low-frequency business data in the low-frequency business data set into a basic threshold autoregressive interval model to obtain an output residual result; adjusting the parameters of the basic threshold autoregressive interval model using the output residual result and the standard residual result to obtain a threshold autoregressive interval model.

[0060] In the above process, the present application trains a threshold autoregressive interval model through low-frequency gasoline business data samples to accurately predict the residual of low-frequency gasoline business data.

[0061] Specifically, the binary empirical mode decomposition (BEMD) is used to decompose the complex-valued signal into a finite number of interval decompositions. Using binary empirical mode decomposition (BEMD) to decompose the complex-valued signal into a finite number of complex-valued intrinsic mode function (IMF) components and residual components, multiple business data fluctuation curves of different frequencies can be obtained, as follows: Binary Empirical Mode Decomposition in Complex Space Therefore, the present application uses BEMD in this study to process the interval value data of gasoline weekly price, including lower and upper bounds.

[0062] To apply the BEMD method, we use the complex-valued signal constructed above. Given a time series of gasoline price intervals , t = 1,2,……,n, this application will The data is mapped to Specifically, this application has As the real part and As the imaginary part, that is:

[0063] here, Next, this application provides a brief introduction to BEMD. The main steps of the BEMD algorithm are as follows: Given a complex-valued signal:

[0064] Step 1: Order

[0065] Step 2: Along (Signal radiation direction) Directionally projecting complex-valued signals

[0066]

[0067] in, represents the real part of a complex-valued signal, = represents the projection direction, M represents the number of projection directions, Indicates the plural form of a unit.

[0068] Step 3: Find out The position corresponding to the local maximum

[0069] Step 4: Use spline interpolation to interpolate the set Interpolate to get the envelope curve ).

[0070] Step 5: Get the average value of all envelope curves , and then subtracted from the signal, that is .make , then return to step 2. Repeat this process until Satisfies the physical meaning of intrinsic eigenmode function (IMF).

[0071] Step 6: Record the obtained IMF and Subtract it from

[0072]

[0073] in, Indicates the first IMF and Represents the residual.

[0074] Step 7: For the remaining residuals Repeat the above steps to obtain all IMFs, that is, from now on, quilt replace.

[0075] Step 8: Get After all the IMFs are included in It can be expressed as: .

[0076] in, represents the total number of IMF, , i = 1,2,..., represents all IMFs, Represents the residual.

[0077] After BEMD, gasoline prices are decomposed into IMF1, IMF2, IMF3, IMF4 and residual. IMF1 is high-frequency data, mainly from short-term spontaneous fluctuations in gasoline prices; IMF2, IMF3 and IMF4 are medium-frequency data reflecting medium-term fluctuations in gasoline prices. Residual is low-frequency data reflecting the long-term trend of gasoline prices.

[0078] Step 120: Input the historical business data in the multiple business data fluctuation curves of different frequencies into the preset multiple business data prediction models to obtain multiple initial business data prediction results and residual results.

[0079] The business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model. The target business data prediction model may be a gasoline price prediction model or the like.

[0080] In some embodiments of the present application, historical business data in multiple business data fluctuation curves of different frequencies are input into multiple preset business data prediction models, including: inputting high-frequency business data and medium-frequency business data in multiple business data fluctuation curves of different frequencies into a target business data prediction model to obtain multiple initial business data prediction results; inputting low-frequency business data in multiple business data fluctuation curves of different frequencies into a threshold autoregressive interval model to obtain residual results.

[0081] In the above process, the present application can accurately predict the business data prediction results of the gasoline to be predicted and the residual results of the gasoline business data respectively through the target business data prediction model (IMLP model) and the threshold autoregressive interval model (TARI model).

[0082] Specifically, for the IMLP part, considering that carbon emission reduction technology can improve engine performance and gasoline combustion efficiency, thereby driving consumer demand, this application compares the prediction accuracy under different standards to determine whether carbon emission reduction technology is helpful for medium-term forecasts. As for TARI, considering that increased carbon concerns may lead to higher carbon emission standards and fuel taxes, which will have a negative impact on natural gas oil supply, this application uses carbon concerns as a threshold variable to capture the nonlinear relationship between carbon concerns and the long-term price components of natural gas oil. The specific operations of the TARI model and the IMLP model are as follows: Threshold Autoregressive Interval Model (TARI): This application applies the Threshold Autoregressive Interval Model (TARI) to medium-frequency and low-frequency components through the method of this application. The TARI model is used to capture the nonlinear characteristics of interval-valued time series and is an interval extension of the classic TAR model, which is designed for point-valued time series. It directly models the interval data and treats each interval observation as a set, rather than modeling the left and right boundaries of the point value separately. In this way, an estimator can be proposed for the TARI model, with the goal of minimizing the distance between the interval model and the interval data. By making more effective use of interval data, this application can achieve asymptotically more efficient inference. In addition, the TARI model can also capture important nonlinear characteristics, such as asymmetric reactions to news and asymmetric level effects. In addition, studies have shown that the out-of-sample prediction results of the TARI model are significantly better than most existing interval models, such as the Linear Autoregressive Conditional Interval (ACI) model and the CR-SETAR model.

[0083] The TARI model can be written as follows:

[0084] in, , represents the interval-valued residual, represents an endogenous or exogenous variable with a point value, p is the lag order used in the autoregression in this application, is the threshold parameter.

[0085] The coefficients are estimated with the minimum Dk distance as the goal:

[0086] Interval Multilayer Perceptron Network: This application uses the Interval Multilayer Perceptron Network (IMLP). Artificial Neural Network (ANN) is a widely used time series prediction model. This application can describe the output and input hysteresis where p is the order of lag, The exogenous variables are listed as follows:

[0087] in, is a linear function that depends on the model structure or hidden layers, is the number of nodes in the last hidden layer. The MLP network is trained in MATLAB using the conjugate gradient error minimization method. Comprehensive prediction. This application combines the upper bounds of all components and all prediction results of the upper bounds obtained by IMLP and TARI to obtain the final gasoline price range prediction results.

[0088] Optionally, a single TARI model can also be used for prediction. The TARI model can be written as follows:

[0089]

[0090] represents the interval-valued residual, represents an endogenous or exogenous variable with a point value, p is the lag order used for autoregression, T is the lag order used for autoregression, is the threshold parameter. The distance is the target for estimating the coefficients.

[0091] .

[0092] Step 130: Determine the final business data prediction result of the gasoline to be predicted through multiple initial business data prediction results and residual results.

[0093] In some embodiments of the present application, a final business data prediction result of the gasoline to be predicted is determined through multiple initial business data prediction results and residual results, including: adding multiple initial business data prediction results and residual results to obtain the final business data prediction result, wherein the business data prediction result includes: gasoline price trend forecast, demand forecast, sales data analysis or production data.

[0094] In the above process, the present application can accurately obtain the final prediction result of the gasoline business data by superimposing the prediction results of the target business data prediction model and the threshold autoregressive interval model.

[0095] Specifically, the prediction results corresponding to different periods are different. The prediction results of gasoline business data corresponding to different periods can be obtained through the following Tables 1 to 6: Table 1 IMF1 prediction results of the IMLP model combining patents and Google trends

[0096] Table 2 Significance test of IMF1 forecast model

[0097] Among them, Table 1 and Table 2 are short-term prediction results. It can be seen from Table 1 and Table 2 that the prediction error of the model with carbon emission reduction technology and carbon concern as variables is greater than that of the model without variables, indicating that carbon emission reduction technology and carbon concern have no impact on gasoline prices in the short term.

[0098] Table 3 Prediction results of IMF2, IMF3 and IMF4 of IMLP model using patents and Google

[0099] Table 4 Significance test of IMF2, IMF3 and IMF4 forecast models

[0100] Medium-term forecast results: For IMF2, IMF3 and IMF4, which adopt carbon emission reduction technology patents and carbon emission Google Trends time series data, the forecast results of this application are shown in Tables 3 and 4. It is worth noting that compared with carbon concerns, incorporating carbon emission reduction technology as an exogenous variable into most criteria of IMF2, IMF3 and IMF4 more significantly improves the forecast performance. In the Friedman test and DM test, it is also seen that the p-value is significant in most comparisons. Therefore, this application continues to incorporate patent time series into IMF2, IMF3 and IMF4 to improve the accuracy of the forecast. The introduction of carbon emission reduction technology patent data can indeed improve the forecast effect of the original model, and the advancement of carbon emission reduction technology has a discernible impact on the medium-term pricing dynamics of gasoline.

[0101] Table 5 Residual prediction results of TARI model for patents and carbon emissions

[0102] Table 6 Significance test of residual prediction model

[0103] Long-term prediction results: For the low-frequency residuals, by adding the threshold Google Trends (i.e., low-carbon economy) as a measure of global carbon emission concerns, as shown in Tables 5 and 6, the present application obtains a huge improvement in significant p-values ​​under almost every criterion of the DM test and Friedman test.

[0104] Optionally, the present application can compare the results predicted by other existing models and the model of the present application by applying TARI, ACI and random forest methods. The specific results can be obtained through the following Tables 7 and 8.

[0105] Table 7. Comparison of models using 20-week, 40-week, and 60-week test data

[0106] Table 8 Significance tests of our model and the comparison model in the test data of 20 weeks, 40 weeks and 60 weeks

[0107] As can be seen from Tables 7 and 8, the model error is basically smaller than that of other models, which shows that our model outperforms other models under almost all criteria, and the p-value is significant compared with the prediction results of other models, which means that our TFIDE learning method can better predict the interval gasoline price, and the global carbon emission problem and carbon reduction technology can significantly affect the gasoline price.

[0108] In the above Figure 1 In the process shown, the present application decomposes the historical business data of gasoline to be predicted by a binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies; inputs the historical business data in the multiple business data fluctuation curves of different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model; through multiple initial business data prediction results and residual results, the final business data prediction result of the gasoline to be predicted is determined. The captured gasoline business data is decomposed by a binary empirical mode decomposition method, and the interval value data is divided into signals of different frequencies for analysis. According to the frequency characteristics of different frequency intervals, a suitable model is selected for prediction, thereby avoiding the loss of volatility information due to the complexity of the overall time series, improving the prediction accuracy and capturing more variability and uncertainty information, greatly improving the out-of-sample prediction of gasoline business data, and achieving the effect of accurately predicting gasoline business data.

[0109] Previous article Figure 1 The following describes the method of interval prediction of gasoline business data. The following takes gasoline price related data as an example. Figure 2 The implementation method of interval prediction of gasoline price in the embodiment of the present application is described in detail.

[0110] Please see Figure 2 , Figure 2 A flowchart of an implementation method for interval prediction of gasoline prices provided in an embodiment of the present application is shown in FIG. Figure 2 The implementation method for the interval forecasted gasoline price shown includes: The process consists of three parts: acquisition and processing of training data, model selection, integration of training and prediction results, and comparison of price predictions with actual prediction results.

[0111] Specifically, the acquisition and processing of training data: constructing a weekly price data set by obtaining daily gasoline prices from the database, obtaining a carbon emission focus data set from Coogle Trends (network), and obtaining emission reduction technology related data from the IncoPal (a database) database.

[0112] Model selection, training and prediction result integration: The video-based interval decomposition gasoline price prediction first decomposes the data through binary empirical mode decomposition, and then processes the data through the preset interval multilayer perceptron network and threshold autoregressive interval to obtain the final gasoline price prediction result.

[0113] Comparison between price prediction and actual prediction results: short-term prediction results are not significant, medium-term prediction results are significant, long-term prediction results are significant, and compared with other models, the results are better than other models.

[0114] also, Figure 2 The specific methods and steps shown can be found in Figure 1 The method shown will not be described in detail here.

[0115] Previous article Figure 1-Figure 2 The method of interval prediction of gasoline business data is described below. Figure 3-Figure 4 A device for interval forecasting gasoline business data is described.

[0116] Please refer to Figure 3 , is a schematic block diagram of a device 300 for predicting gasoline business data within an interval provided in an embodiment of the present application. The device 300 may be a module, program segment or code on an electronic device. The device 300 is similar to the above Figure 1 The method embodiment corresponds to and can be executed Figure 1 The various steps involved in the method embodiment and the specific functions of the device 300 can be found in the description below. To avoid repetition, the detailed description is appropriately omitted here.

[0117] Optionally, the device 300 includes: A decomposition module 310 is used to decompose the historical business data of gasoline to be predicted by using a binary empirical mode decomposition method to obtain a plurality of business data fluctuation curves of different frequencies; The prediction module 320 is used to input the historical business data in the business data fluctuation curves of multiple different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model; The determination module 330 is used to determine the final business data prediction result of the gasoline to be predicted through multiple initial business data prediction results and residual results.

[0118] Optionally, the device further comprises: A training module is used for the decomposition module to decompose the historical business data of the gasoline to be predicted by the binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies, collect all historical business data of various types of gasoline, and decompose all historical business data by the binary empirical mode decomposition method to obtain a set of all historical business data; classify the historical business data in the set of all historical business data according to preset frequencies to obtain multiple classification sets, wherein the classification sets include high-frequency business data sets, medium-frequency business data sets or low-frequency business data sets; train the target business data prediction model by the high-frequency business data sets and medium-frequency business data sets in the multiple classification sets; and train the threshold autoregressive interval model by the low-frequency business data sets in the multiple classification sets.

[0119] Optionally, the training module is specifically used to: Add exogenous variables to the high-frequency business data set and the medium-frequency business data set to obtain a modified high-frequency business data set and a modified medium-frequency business data set, wherein the exogenous variables represent the standard carbon emission data set set in the carbon emission reduction technology; input the modified high-frequency business data set and the modified medium-frequency business data set into the basic data prediction model to obtain prediction results; adjust the parameters of the basic data prediction model through the prediction results to obtain the target business data prediction model.

[0120] Optionally, the training module is specifically used to: The low-frequency service data in the low-frequency service data set is input into the basic threshold autoregressive interval model to obtain an output residual result; the parameters of the basic threshold autoregressive interval model are adjusted by the output residual result and the standard residual result to obtain the threshold autoregressive interval model.

[0121] Optionally, the prediction module 320 is specifically used for: The high-frequency business data and medium-frequency business data in multiple business data fluctuation curves with different frequencies are input into the target business data prediction model to obtain multiple initial business data prediction results; the low-frequency business data in multiple business data fluctuation curves with different frequencies are input into the threshold autoregressive interval model to obtain the residual results.

[0122] Optionally, the determination module 330 is specifically configured to: Multiple initial business data prediction results and residual results are added together to obtain a final business data prediction result, wherein the business data prediction result includes: gasoline price trend prediction, demand forecast, sales data analysis or production data.

[0123] Please refer to Figure 44 is a schematic block diagram of a device for predicting gasoline business data within a certain interval provided in an embodiment of the present application. The device may include a memory 410 and a processor 420. Optionally, the device may also include: a communication interface 430 and a communication bus 440. The device is connected to the above Figure 1 The method embodiment corresponds to and can be executed Figure 1 The various steps involved in the method embodiment and the specific functions of the device can be found in the description below.

[0124] Specifically, the memory 410 is used to store computer-readable instructions.

[0125] Processor 420 is used to process the readable instructions stored in the memory and can execute Figure 1 The steps in the method.

[0126] The communication interface 430 is used for signaling or data communication with other node devices, for example, for communication with a server or a terminal, or for communication with other device nodes, but the embodiments of the present application are not limited thereto.

[0127] The communication bus 440 is used to realize direct connection and communication among the above components.

[0128] The communication interface 430 of the device in the embodiment of the present application is used to communicate with other node devices for signaling or data. The memory 410 can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The memory 410 can also be at least one storage device located away from the aforementioned processor. The memory 410 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 420, the electronic device executes the aforementioned Figure 1 The method process shown. The processor 420 can be used on the device 300 and used to perform the functions in the present application. Exemplarily, the above-mentioned processor 420 can be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and the embodiments of the present application are not limited thereto.

[0129] The embodiment of the present application also provides a readable storage medium, when the computer program is executed by a processor, Figure 1 The method process in the method embodiment shown is executed by the electronic device.

[0130] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method, and will not be described in detail here.

[0131] In summary, the embodiment of the present application provides a method, device, equipment and storage medium for interval prediction of gasoline business data, the method comprising: decomposing the historical business data of gasoline to be predicted by a binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies; inputting the historical business data in the multiple business data fluctuation curves of different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model comprises: a threshold autoregressive interval model or a target business data prediction model; determining the final business data prediction result of the gasoline to be predicted by multiple initial business data prediction results and residual results. This method can achieve the effect of accurately predicting gasoline business data.

[0132] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0133] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0134] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0135] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0136] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0137] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A method for interval prediction of gasoline business data, characterized in that: include: The historical business data of gasoline to be predicted is decomposed by the binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies; Inputting the historical business data in the multiple business data fluctuation curves of different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model; The final business data prediction result of the gasoline to be predicted is determined through the multiple initial business data prediction results and residual results.

2. The method according to claim 1, characterized in that Before decomposing the historical business data of gasoline to be predicted by the binary empirical mode decomposition method to obtain a plurality of business data fluctuation curves of different frequencies, the method further includes: Collect all historical business data of various types of gasoline, and decompose all the historical business data by using the binary empirical mode decomposition method to obtain a set of all historical business data; Classifying the historical business data in the entire historical business data set according to a preset frequency to obtain a plurality of classification sets, wherein the classification sets include a high-frequency business data set, a medium-frequency business data set, or a low-frequency business data set; Training the target service data prediction model by using the high-frequency service data set and the medium-frequency service data set in the multiple classification sets; The threshold autoregressive interval model is trained by using the low-frequency service data set in the multiple classification sets.

3. The method according to claim 2, characterized in that The training of the target service data prediction model by using the high-frequency service data set and the medium-frequency service data set in the multiple classification sets includes: Adding exogenous variables to the high-frequency business data set and the medium-frequency business data set to obtain a modified high-frequency business data set and a modified medium-frequency business data set, wherein the exogenous variables represent standard carbon emission data set in the carbon emission reduction technology; Inputting the modified high-frequency service data set and the modified medium-frequency service data set into a basic data prediction model to obtain a prediction result; The parameters of the basic data prediction model are adjusted according to the prediction results to obtain the target business data prediction model.

4. The method according to claim 2, characterized in that: The step of training the threshold autoregressive interval model by using the low-frequency service data set in the multiple classification sets includes: Inputting the low-frequency service data in the low-frequency service data set into a basic threshold autoregressive interval model to obtain an output residual result; The parameters of the basic threshold autoregressive interval model are adjusted by using the output residual result and the standard residual result to obtain the threshold autoregressive interval model.

5. The method according to any one of claims 1 to 4, characterized in that: The step of inputting the historical business data in the multiple business data fluctuation curves of different frequencies into the preset multiple business data prediction models includes: Inputting the high-frequency business data and the medium-frequency business data in the multiple business data fluctuation curves of different frequencies into the target business data prediction model to obtain the multiple initial business data prediction results; The low-frequency business data in the multiple business data fluctuation curves with different frequencies are input into the threshold autoregressive interval model to obtain the residual result.

6. The method according to any one of claims 1 to 4, characterized in that: The step of determining the final business data prediction result of the gasoline to be predicted by using the multiple initial business data prediction results and residual results includes: The multiple initial business data prediction results and the residual results are added together to obtain the final business data prediction result, wherein the business data prediction result includes: gasoline price trend prediction, demand forecast, sales data analysis or production data.

7. A device for interval prediction of gasoline business data, characterized in that: include: A decomposition module is used to decompose the historical business data of gasoline to be predicted by using a binary empirical mode decomposition method to obtain multiple business data fluctuation curves of different frequencies; A prediction module, used for inputting the historical business data in the multiple business data fluctuation curves of different frequencies into multiple preset business data prediction models to obtain multiple initial business data prediction results and residual results, wherein the business data prediction model includes: a threshold autoregressive interval model or a target business data prediction model; A determination module is used to determine the final business data prediction result of the gasoline to be predicted through the multiple initial business data prediction results and residual results.

8. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method according to any one of claims 1 to 6 are executed.

9. A computer-readable storage medium, characterized in that: include: A computer program, when the computer program is run on a computer, causes the computer to execute the method according to any one of claims 1 to 6.

Images