CN114066004A - GDP (graphics device performance) current prediction method based on electric power big data - Google Patents

GDP (graphics device performance) current prediction method based on electric power big data Download PDF

Info

Publication number
CN114066004A
CN114066004A CN202111162746.9A CN202111162746A CN114066004A CN 114066004 A CN114066004 A CN 114066004A CN 202111162746 A CN202111162746 A CN 202111162746A CN 114066004 A CN114066004 A CN 114066004A
Authority
CN
China
Prior art keywords
prediction
model
data
midas
gdp
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.)
Pending
Application number
CN202111162746.9A
Other languages
Chinese (zh)
Inventor
彭放
孙妮
肖娅晨
潘璠
祁亚茹
刘甜甜
任俊达
刘化龙
李博
王敏楠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Big Data Center Of State Grid Corp Of China
Original Assignee
Big Data Center Of State Grid Corp Of China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Big Data Center Of State Grid Corp Of China filed Critical Big Data Center Of State Grid Corp Of China
Priority to CN202111162746.9A priority Critical patent/CN114066004A/en
Publication of CN114066004A publication Critical patent/CN114066004A/en
Pending legal-status Critical Current

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
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2462—Approximate or statistical queries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Development Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Probability & Statistics with Applications (AREA)
  • Operations Research (AREA)
  • Mathematical Physics (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Databases & Information Systems (AREA)
  • Educational Administration (AREA)
  • Health & Medical Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Game Theory and Decision Science (AREA)
  • Mathematical Optimization (AREA)
  • Software Systems (AREA)
  • Mathematical Analysis (AREA)
  • Quality & Reliability (AREA)
  • Computational Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Fuzzy Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Algebra (AREA)
  • Artificial Intelligence (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)

Abstract

The invention provides a GDP (graphics device performance) current prediction method based on electric power big data, which comprises the following steps: constructing a plurality of GDP current prediction models based on historical sample data, wherein the historical sample data comprises economic statistical data and/or power data of a historical time period; training a plurality of GDP current prediction models by using different prediction methods based on historical sample data to obtain a plurality of combined mean square prediction errors of the different prediction methods and the plurality of GDP current prediction models; determining an optimal combination according to the mean square prediction errors of all combinations; performing current prediction on the GDP by using the optimal combination and economic statistical data and/or electric power data of a preset time period; the different prediction methods include a mixed data sampling model method, a prediction combining method, and a dynamic factor method. The invention can predict GDP more timely and accurately.

Description

GDP (graphics device performance) current prediction method based on electric power big data
Technical Field
The invention relates to the technical field of GDP current prediction based on electric power big data, in particular to a GDP current prediction method based on electric power big data.
Background
Whether government economic policies are established or enterprise operation policies are selected, the current macroscopic economic situation needs to be known timely and accurately. GDP is the most core economic indicator for describing macroscopic economic situation, and has been paid much attention from various economic subjects. Although the prediction of GDP has been the focus of many researchers and institutional research, the timeliness and accuracy of GDP prediction are limited by the shortcomings of conventional prediction techniques and the lag of common economic statistics.
Along with the human society stepping into the big data era, the macro-economic current prediction based on big data receives more and more attention. The electric power big data is closely related to economic life and is a 'weather meter' for national economic operation, the characteristics of high real-time frequency, objectivity, accuracy, fine granularity and the like enable the electric power big data to be used for forecasting the macro economic situation at present, but a method for effectively utilizing the electric power big data to forecast the macro economic situation at present is lacked.
Disclosure of Invention
The invention provides a GDP (graphics data processing) current prediction method based on big electric data, which is used for realizing the GDP current prediction method based on the big electric data and aims to solve the problems that the timeliness of explanatory variable data is weak and valuable information cannot be fully utilized in the existing GDP prediction based on the big electric data, a metrological and economics mixing technology, a big data dimension reduction technology, a prediction combination technology and the like so as to realize more timely and accurate prediction of GDP.
The invention provides a GDP (graphics device performance) current prediction method based on electric power big data, which comprises the following steps:
step 1, constructing a plurality of GDP current prediction models based on historical sample data, wherein the historical sample data comprises economic statistical data and/or power data of a historical time period;
step 2, training a plurality of GDP current prediction models by using different prediction methods based on historical sample data to obtain a plurality of combined mean square prediction errors of the different prediction methods and the plurality of GDP current prediction models;
step 3, determining an optimal combination according to the mean square prediction errors of all combinations;
step 4, performing current prediction on the GDP by using the optimal combination and economic statistical data and/or electric power data of a preset time period;
the different prediction methods include a mixed data sampling model method, a prediction combining method, and a dynamic factor method.
Preferably, the GDP current prediction model includes a first model, a second model, a third model, a fourth model and a fifth model.
Preferentially, the first model is used for predicting the Tth period value of the prediction target yTth period by utilizing traditional economic statistical data; calculating and obtaining a first model of the predictive model by equation (1):
Figure BDA0003290754820000021
wherein the content of the first and second substances,
Figure BDA0003290754820000022
an information set of statistical data for the prediction;
Figure BDA0003290754820000023
predicting the value of the used ith economic statistic interpretation variable at the t stage; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
hxto predict the step size, when hxWhen the current prediction is equal to 1 or 2, the information of 2 months before the current period and 1 month before the current period are respectively called for prediction.
Preferably, the second model is used for predicting the Tth period value of the prediction target y by using traditional economic statistical data and power data; during prediction, statistical data prediction is used as a main part, and power data prediction is used as an auxiliary part; calculating and obtaining a second model of the predictive model by equation (2):
Figure BDA0003290754820000031
wherein the content of the first and second substances,
Figure BDA0003290754820000032
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure BDA0003290754820000033
representing a set of power data information used in the prediction;
Figure BDA0003290754820000034
the used ith power data index; m represents the frequency multiplication difference, epsilon, of the explanatory variable and the predicted target variableTRepresenting a prediction residual, T representing a period of a prediction target value to be predicted;
Figure BDA0003290754820000035
this indicates that prediction is performed using information that is contained in the power data but not in the statistical data.
Preferably, the third model is used for predicting the Tth period value of the prediction target y by using the power data; calculating and obtaining a third model of the prediction model by equation (3):
Figure BDA0003290754820000036
predicting the Tth period value of the prediction target y by using the power data; in the model
Figure BDA0003290754820000037
Representing a set of power data information used in the prediction;
Figure BDA0003290754820000038
Figure BDA0003290754820000039
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
when predicting step length hzEqual to 0, 1 or 2 is the current prediction, and the information of the current 3 months, the previous 2 months and the previous 1 month is used for prediction.
Preferably, the fourth model predicts the value of the prediction target y during the Tth period by using traditional economic statistical data and power data, and the prediction is mainly performed on the power data and is assisted by the statistical data; calculating and obtaining a fourth model of the predictive model by equation (4):
Figure BDA0003290754820000041
wherein the content of the first and second substances,
Figure BDA0003290754820000042
an information set of statistical data for the prediction;
Figure BDA0003290754820000043
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure BDA0003290754820000044
representing a set of power data information used in the prediction;
Figure BDA0003290754820000045
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
Figure BDA0003290754820000046
this indicates that prediction is performed using information that is included in the statistical data but not included in the power data.
Preferably, the fifth model is used for predicting the Tth period value of the prediction target y by using the power data; a fifth model of the prediction model is calculated and obtained by equation (5):
Figure BDA0003290754820000047
wherein the content of the first and second substances,
Figure BDA0003290754820000048
an information set of statistical data for the prediction;
Figure BDA0003290754820000049
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure BDA00032907548200000410
represents a set of power data information used in the prediction,
Figure BDA00032907548200000411
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t denotes the period of time in which the prediction target value to be predicted is present.
Preferably, in step 2, based on historical sample data, the method for sampling frequency mixing data is used to train a plurality of GDP current prediction models respectively, so as to obtain a mean square prediction error of a plurality of combinations of the method for sampling frequency mixing data and the plurality of GDP current prediction models, including:
training N combinations formed by any one mixing data sampling model and N GDP current prediction models based on historical sample data, and obtaining mean square prediction errors of the N combinations;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model or an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
Preferably, in step 2, based on historical sample data, the plurality of GDP current prediction models are respectively trained by using a prediction combination method to obtain a mean square prediction error of a plurality of combinations of the prediction combination method and the plurality of GDP current prediction models, where the mean square prediction error includes:
respectively training M multiplied by N combinations formed by M mixing data sampling models and N prediction models based on historical sample data, and obtaining mean square prediction errors of the M multiplied by N combinations;
calculating the predicted values of the M multiplied by N combinations of the frequency mixing data sampling model method and the prediction model by using a prediction combination method, and calculating the mean square prediction error of the prediction combination;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model and an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
Preferably, in step 2, based on historical sample data, the dynamic factor method is used to train the multiple GDP current prediction models respectively, so as to obtain multiple combinations of mean square prediction errors of the dynamic factor method and the multiple GDP current prediction models, where the multiple combinations of the dynamic factor method and the multiple GDP current prediction models are combined, and the method includes:
calculating a common factor by using the dynamic factor model;
respectively training N combinations formed by any one of the mixing data sampling models and N prediction models by taking the common factor as a new high-frequency interpretation variable, and obtaining the mean square prediction error of the N combinations;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model or an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
Preferably, the determining an optimal combination according to the mean square prediction errors of all combinations includes:
selecting a combination with the minimum mean square prediction error from all combinations as an optimal combination;
and comparing the mean square prediction error of the optimal combination with the mean square prediction error of the prediction combination method, wherein if the mean square prediction error of the optimal combination is less than the mean square prediction error of the prediction combination method, the optimal combination is the optimal scheme for GDP current prediction, and otherwise, the prediction combination method is the optimal scheme for GDP current prediction.
The technical scheme of the invention has the following beneficial effects:
the invention provides a GDP (global data projection) current prediction method based on big electric data, which is mainly based on historical sample data and is used for training a plurality of GDP current prediction models by using different prediction methods to obtain the mean square prediction errors of various combinations of the different prediction methods and the plurality of GDP current prediction models; and determining the optimal combination according to the mean square prediction errors of all combinations. The GDP is predicted more timely and accurately at present through the optimal combination obtained by the technical scheme.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
Fig. 1 is a flowchart illustrating a GDP current prediction method according to an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
Currently, the GDP prediction in China mainly adopts a common-frequency prediction technology in the metrological economics, and the used explanatory variables are mainly selected economic statistical indexes. GDP data issued by the national statistical bureau is seasonal data, and the GDP data predicted by applying the same-frequency prediction technology needs to be added with high-frequency explanatory variable data into low-frequency data, which causes some important information to be lost, and influences the accuracy and timeliness of prediction. The same-frequency prediction technology is difficult to process large-scale data problems and can only be carried out by utilizing a plurality of selected economic statistical variables. The current macroscopic economy transmission mechanism is more complex, GDP influence factors are more diversified, and prediction based on the same-frequency technology cannot fully utilize a large amount of effective information, particularly current information about economic situations, so as to improve the accuracy of GDP prediction.
Example 1
Based on this, an embodiment of the present invention provides a GDP current prediction method based on big power data, as shown in fig. 1, including the following steps:
step 1, constructing a plurality of GDP current prediction models based on historical sample data, wherein the historical sample data comprises economic statistical data and/or power data of a historical time period;
step 2, training a plurality of GDP current prediction models by using different prediction methods based on historical sample data to obtain a plurality of combined mean square prediction errors of the different prediction methods and the plurality of GDP current prediction models;
step 3, determining an optimal combination according to the mean square prediction errors of all combinations;
step 4, performing current prediction on the GDP by using the optimal combination and economic statistical data and/or electric power data of a preset time period;
the different prediction methods may include a mixed data sampling model method, a prediction combining method, and a dynamic factor method, among others.
In step 1, the data used in the embodiment of the present invention may include two types: power data and traditional economic statistics;
the power data may include various power data indicators. The used power data is monthly electricity consumption same-proportion growth rate obtained by summing up 26 electricity consumption data of 5 hundred million users in China, and comprises 90 indexes including company whole-industry total electricity consumption same-proportion growth rate, urban and rural resident total electricity consumption same-proportion growth rate, whole-industry total electricity consumption same-proportion growth rate and whole-industry total electricity consumption same-proportion growth rate of each subdivision industry under different divisions;
the traditional statistical data mainly comprises 42 variables such as social consumer goods retail total amount, fixed generation investment total amount, export total value, nominal effective exchange rate, various indexes of a certified exchange and a deep-certified exchange, business confidence index, trade index, currency supply, deposit, macroscopic economy landscape index and the like, and the data are derived from a CEIC database.
The GDP current prediction method based on the big power data aims to solve the problems that interpretation variable data in the existing GDP prediction is poor in timeliness and valuable information cannot be fully utilized based on the big power data, a metrological and economics mixing technology, a big data dimension reduction technology, a prediction combination technology and the like so as to achieve more timely and accurate prediction of the GDP.
Furthermore, in order to fully utilize the information of economic statistical data and electric power big data to obtain a more timely and accurate prediction result, five modes of combining two types of data are designed, and the following five types of prediction models are correspondingly constructed; specifically, the method comprises the following steps:
the predictive models include a first model, a second model, a third model, a fourth model, and a fifth model.
The first model is:
predicting the value of the predicted target y at the T stage by using traditional economic statistical data; calculating and obtaining a first model of the predictive model by equation (1):
Figure BDA0003290754820000091
wherein the content of the first and second substances,
Figure BDA0003290754820000092
an information set of statistical data for the prediction;
Figure BDA0003290754820000093
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
hxto predict the step size, when hxWhen the current prediction is equal to 1 or 2, the information of 2 months before the current period and 1 month before the current period are respectively called for prediction.
The second model is:
predicting the value of the Tth period of the prediction target yI by using traditional economic statistical data and power data; during prediction, statistical data prediction is used as a main part, and power data prediction is used as an auxiliary part; calculating and obtaining a second model of the predictive model by equation (2):
Figure BDA0003290754820000094
Figure BDA0003290754820000101
wherein the content of the first and second substances,
Figure BDA0003290754820000102
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure BDA0003290754820000103
representing a set of power data information used in the prediction;
Figure BDA0003290754820000104
the used ith power data index; m represents the frequency multiplication difference, epsilon, of the explanatory variable and the predicted target variableTRepresenting a prediction residual, T representing a period of a prediction target value to be predicted;
Figure BDA0003290754820000105
this indicates that prediction is performed using information that is contained in the power data but not in the statistical data.
Since the distribution of the economic statistic data generally lags behind one month, the information used for prediction is one month less than that of the power data.
The third model is:
predicting the Tth period value of the prediction target y by using the power data; calculating and obtaining a third model of the prediction model by equation (3):
Figure BDA0003290754820000106
and predicting the Tth period value of the prediction target y by using the power data. In the model
Figure BDA0003290754820000107
Representing a set of power data information used in the prediction;
Figure BDA0003290754820000108
Figure BDA0003290754820000109
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present; when predicting step length hzEqual to 0, 1 or 2 is the current prediction, and the information of the current 3 months, the previous 2 months and the previous 1 month is used for prediction.
The fourth model is:
in order to predict the value of the prediction target y at the Tth period by using traditional economic statistical data and power data, the prediction of the power data is taken as the main part during prediction, and the prediction of the statistical data is taken as the auxiliary part; calculating and obtaining a fourth model of the predictive model by equation (4):
Figure BDA0003290754820000111
wherein the content of the first and second substances,
Figure BDA0003290754820000112
an information set of statistical data for the prediction;
Figure BDA0003290754820000113
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure BDA0003290754820000114
representing a set of power data information used in the prediction;
Figure BDA0003290754820000115
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
Figure BDA0003290754820000116
this indicates that prediction is performed using information that is included in the statistical data but not included in the power data.
The fifth model is:
predicting the Tth period value of the prediction target y by using the power data; a fifth model of the prediction model is calculated and obtained by equation (5):
Figure BDA0003290754820000117
Figure BDA0003290754820000118
an information set of statistical data for the prediction;
Figure BDA0003290754820000119
Figure BDA00032907548200001110
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure BDA00032907548200001111
represents a set of power data information used in the prediction,
Figure BDA00032907548200001112
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t denotes the period of time in which the prediction target value to be predicted is present.
The fifth model comprehensively utilizes traditional statistical data and power data to predict the value of the predicted target y in the Tth period, but the fifth model is different from the second model and the fourth model in that the fifth model treats two types of data without difference in prediction.
F () in the above-described first to fifth model formulas is a prediction method used.
In the step 2, for the five types of prediction models, the current prediction of the GDP may be performed by comprehensively using three methods, namely, a mixing data sampling Model (MIDAS), a dynamic factor and a prediction combination, as an example, which is described in detail below; specifically, the method comprises the following steps:
step 21, performing prediction training by using a mixed frequency data sampling Model (MIDAS) method
In the embodiment of the present invention, based on historical sample data, a frequency mixing data sampling model method is used to train a plurality of GDP current prediction models, so as to obtain a mean square prediction error of a plurality of combinations of the frequency mixing data sampling model method and the plurality of GDP current prediction models, which may include the following steps:
training N combinations formed by any one mixing data sampling model and N GDP current prediction models based on historical sample data, and obtaining mean square prediction errors of the N combinations;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model or an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
Step 22, performing prediction training by using a prediction combination model method
In the embodiment of the present invention, based on historical sample data, a prediction combination method is used to train a plurality of GDP current prediction models, so as to obtain a mean square prediction error of a plurality of combinations of the prediction combination method and the plurality of GDP current prediction models, which may include the following steps:
respectively training M multiplied by N combinations formed by M mixing data sampling models and N prediction models based on historical sample data, and obtaining mean square prediction errors of the M multiplied by N combinations;
calculating the predicted values of the M multiplied by N combinations of the frequency mixing data sampling model method and the prediction model by using a prediction combination method, and calculating the mean square prediction error of the prediction combination;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model and an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
Step 23, performing prediction training by using dynamic factor model method
In the embodiment of the present invention, based on historical sample data, a dynamic factor method is used to train a plurality of GDP current prediction models, so as to obtain a plurality of combinations of mean square prediction errors of the dynamic factor method and the plurality of GDP current prediction models, which may include the following steps:
calculating a common factor by using the dynamic factor model;
respectively training N combinations formed by any one of the mixing data sampling models and N prediction models by taking the common factor as a new high-frequency interpretation variable, and obtaining the mean square prediction error of the N combinations;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model or an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
In the embodiment of the present invention, there is no sequence for the execution of the above steps 21 to 23.
In the step 3, the optimal combination/optimal scheme may be determined by any one of the following methods:
1) the method comprises the following steps:
and directly selecting the combination with the minimum mean square prediction error from all combinations as the optimal combination.
2) Method two
Selecting a combination with the minimum mean square prediction error from all combinations as an optimal combination;
and comparing the mean square prediction error of the optimal combination with the mean square prediction error of the prediction combination method, wherein if the mean square prediction error of the optimal combination is less than the mean square prediction error of the prediction combination method, the optimal combination is the optimal scheme for GDP current prediction, and otherwise, the prediction combination method is the optimal scheme for GDP current prediction.
3) Method III
Selecting a combination with the minimum mean square prediction error from all combinations;
comparing the mean square prediction error of the combination with the mean square prediction error of the reference combination, and if the mean square prediction error of the combination is less than the mean square prediction error of the reference combination, taking the combination as an optimal combination; otherwise, the reference combination is used as the optimal combination;
the mean square prediction error of the reference combination is that a mixing data sampling model method, a prediction combination method and dynamic factors are used for predicting the reference model respectively, and a method corresponding to the minimum value in the mean square prediction errors obtained by prediction is selected to form the reference combination together with the reference model.
The specific steps in step 4 may include the following steps:
acquiring economic statistical data and/or electric power data of a preset time period according to the GDP current prediction model in the optimal combination determined in the step 3;
for example, if the GDP current prediction model corresponding to the optimal combination is constructed by using the power data, the GDP current prediction is performed by using the optimal combination based on the acquired power data of the preset time period;
if the GDP current prediction model corresponding to the optimal combination is constructed by using economic statistical data, current prediction is carried out on the GDP by using the optimal combination based on the obtained economic statistical data of the preset time period;
and if the GDP current prediction model corresponding to the optimal combination is constructed by using the economic statistical data and the electric power data, performing current prediction on the GDP by using the optimal combination based on the obtained economic statistical data and the electric power data in the preset time period.
Example 2
In step 21 of embodiment 1 of the present invention, a mixed data sampling Model (MIDAS) is used for prediction training, and any one of the following models may be used: and (3) predicting by using a univariate MIDAS model:
constructing a univariate MIDAS model for comparing data of a prediction target of the power data; the main characteristic of constructing the univariate MIDAS model is that the univariate MIDAS model passes through a weight function B (L)1/m(ii) a Theta) converting high frequency data
Figure BDA0003290754820000151
The direct introduction of the model avoids the low frequency processing of high frequency data.
The calculation is performed according to equation (6):
Figure BDA0003290754820000152
m represents the frequency multiplication difference between the high-frequency data and the low-frequency data;
Figure BDA0003290754820000153
Figure BDA0003290754820000154
w (k; θ) is a weighting function; beta is a0Represents a constant term; beta is a1The coefficients representing the weight function B are,
Figure BDA0003290754820000155
representing a high frequency time series; t is 1, …, and T is a low-frequency sequence time unit;
Figure BDA0003290754820000156
a hysteresis operator representing the high frequency data; theta represents a parameter to be estimated;
Figure BDA0003290754820000157
k is the set maximum lag order of the high-frequency data; epsilontRepresenting a random error;
wherein a prediction model MIDAS (m, K, h) of the univariate MIDAS model is represented by formula (7):
Figure BDA0003290754820000158
wherein h represents a prediction step size; y ist+h|tRepresents the prediction of the y value in the t + h period in the t period; if the prediction target is quarterly data, the explanation variable is high-frequency monthly data, and the current prediction is carried out when h is less than 3; when h is equal to 0, 1 and 2, the monthly data of the current period of 3 months, 2 months and 1 month respectively correspond to the current prediction; when predicting GDP, interpreting variables as being extracted from current data; xit+hRepresents the prediction residual; when predicting GDP, interpreting variables as being extracted from current data;
the weight function is obtained by calculation using the exponential almon polynomial function in formula (8), so that the weight required for predicting the target is satisfied, and the weight is positive, and the error of the variance tends to zero: that is, the exponential Almon polynomial parameter constraint can ensure that the weights required for the analysis are satisfied on the one hand, and can ensure that the weights are positive and the variance error approaches zero on the other hand.
Figure BDA0003290754820000161
Wherein, let θ1≤300,θ2<0; exp () represents an exponential function. In the embodiment of the invention, only theta in the formula (8) is selected1And theta2Two parameters.
2) And (3) predicting by using the MIDAS-AR model:
constructing an MIDAS-AR model and predicting macroscopic economic variables in traditional economic statistical data;
inertia generally exists in the macro-economic variables, so that when the macro-economic variables are predicted, a hysteresis term of the macro-economic variables is necessary to be introduced into a univariate MIDAS model as an explanatory variable; the MIDAS-AR model with a p-order autoregressive lag term can be represented by equation (9):
Figure BDA0003290754820000162
wherein λ represents the coefficient to be estimated；yt-pRepresenting a lagged term of order p;
Figure BDA0003290754820000163
representing a high frequency time series; epsilontRepresenting a random error;
obtaining a corresponding prediction model MIDAS (m, K, h) -AR (p) by the MIDAS-AR model, wherein the expression is shown as a formula (10):
Figure BDA0003290754820000164
wherein ξtRepresenting the prediction residual.
3) Prediction using m (n) -MIDAS model:
constructing a corresponding M (n) -MIDAS model based on the univariate MIDAS model and the MIDAS-AR model, and predicting the complex economic data in the macroscopic economic variables;
macroscopic economic variables are in a complex economic system and are influenced by a plurality of economic factors, so that a multivariable MIDAS model is necessary to predict. The expressions of the multivariate models M (n) -MIDAS and M (n) -MIDAS-AR corresponding to MIDAS and MIDAS-AR are calculated according to the formula (11):
Figure BDA0003290754820000171
Figure BDA0003290754820000172
wherein n represents the number of interpretation variables in the model; b isiA polynomial representing the ith weight function; beta is a0Denotes a constant term, βiRepresenting coefficients before the weight function, which are all parameters to be estimated;
Figure BDA0003290754820000173
represents the ith time series; epsilontRepresenting a random error;
wherein n represents the number of interpretation variables in the model;
the prediction models M (n) -MIDAS (m, K, h), M (n) -MIDAS (m, K, h) -AR corresponding to M (n) -MIDAS and M (n) -MIDAS-AR (p) are shown in formula (12):
Figure BDA0003290754820000174
Figure BDA0003290754820000175
wherein ξtRepresenting the prediction residual.
Example 3
In step 22 in embodiment 1 of the present invention, a prediction combination model method is used for prediction training, and the following method may also be adopted:
combining the plurality of prediction models to obtain a prediction combination model,
when the predicted target is influenced by a plurality of interpretation variables, the data sample capacity and excessive parameters to be estimated limit the prediction effect of M (n) -MIDAS. In this case, predictive combination (Forecast combination) is a commonly used solution. By summarizing the prediction combination method, it is considered that the prediction accuracy can be improved by combining the prediction results of several models by a certain weight. The predicted combination of n models at time t can be calculated according to equation (13):
combining the plurality of prediction models and obtaining a prediction combination model, the prediction combination model being calculated according to equation (13):
Figure BDA0003290754820000181
wherein the content of the first and second substances,
Figure BDA0003290754820000182
indicating the ith prediction at time t, wi,tRepresenting the weight of the ith prediction result in the prediction combination;
wherein the weight may be in an equally weighted form; the BIC criteria may also be used as weights, i.e. expressed in equation (14),
Figure BDA0003290754820000183
wherein, BICiAnd the BIC information criterion value represents the ith prediction result, and n represents the number of prediction models.
Example 4
In step 23 in embodiment 1 of the present invention, a dynamic factor model method is used for prediction training, and the following method may be used:
constructing a dynamic factor model for extracting a small amount of common factors containing rich information from high-dimensional data;
the dynamic factor model can extract a small amount of common factors containing rich information from the high-dimensional data, thereby realizing the dimension reduction of the high-dimensional data. The common factors can be used for predicting, constructing indexes and the like, and the dynamic factor model is widely applied to the macro-metering economy in recent ten years; the dynamic factor model is calculated by equation (15):
Xt＝λ(L)ft+et
ft＝A1ft-1+…+Apft-p+ηt (15)；
wherein, Xt(T1, …, T) is an N-dimensional vector, λ (L) is an N × q order lag polynomial matrix, etRepresenting a heterogeneous interference vector; f. oft＝(f1t,…,fqt) Represents a common factor; f. oft-1、ft-pDenotes ft1 st order and p th order lags; etatRepresenting a heterogeneous impact vector; a. thei(i ═ 1, …, p) is a matrix of autoregressive coefficients for the factors;
wherein, let Ft＝(ft′,f′t-1,…,f′t-p) Is an r × 1 order vector, let Λ ═ λ0,λ1,…,λp)；
Wherein λ isiOf lambda (L)A coefficient matrix corresponding to the i-order lag operator is obviously an Nxq-order matrix;
let phi (L) be 0, 1 and Ai(i ═ 1, …, p), the dynamic factor model can be restated as in equation (16):
Xt＝ΛFt+et
Ф(L)Ft＝Gηt (16)；
wherein G is a matrix formed by 0 and 1;
the two above equations are referred to as the static form of the dynamic factor model. Bai & Ng (2007) indicates that there is little difference between the static factor model and the dynamic factor model from a prediction perspective.
The frequency domain estimation method of DFM cannot directly estimate ftTherefore, it cannot be used for prediction (Stock)&Watson, 2011). The time domain estimation method of DFM is mainly described next. Stock&Watson (2011) divides the time domain estimation method of DFM into three generations. The first generation is maximum likelihood estimation based on Kalman filtering, the second generation is a nonparametric averaging method, and the third generation is a method for combining principal components and a state space.
The present invention mainly adopts a principal component estimation method in the second generation method. According to a principal component estimation method, a factor matrix
Figure BDA0003290754820000191
Eigenvectors corresponding to the k largest eigenvalues and
Figure BDA0003290754820000192
by the product of (c), and factor load estimation
Figure BDA0003290754820000193
Further, the estimation of the number q of factors is calculated by equation (17):
the estimation of the number of factors q is calculated by equation (17):
Figure BDA0003290754820000201
wherein the content of the first and second substances,
Figure BDA0003290754820000202
representing factor number estimation values;
Figure BDA0003290754820000203
mean value representing the sum of the squares of the residuals;
Figure BDA0003290754820000204
and
Figure BDA0003290754820000205
respectively representing the factor and the estimated value of the factor load under the given condition of the factor number q, and p (N, T) is a penalty function.
Example 5
In step 3 in embodiment 1 of the present invention, the following method may be adopted to determine the optimal combination:
1. reference model and effect evaluation index
The invention selects ARMA (Autoregressive moving average model) as a reference model, and uses rMSFE (ratio of mean square predicted error ratios) to measure the prediction effects of different models and methods. The mean square prediction error ratio is the mean square prediction error ratio of the non-reference model and the reference prediction model, the ratio is smaller than 1, which indicates that the non-reference model has better prediction effect than the reference model, and the smaller the ratio is, the better the prediction effect of the non-reference model is.
2. Determining optimal power metrics and prediction models
The electric power data indexes are divided into different levels, and the electric power indexes of different levels and GDP prediction effects combined with statistical data are different. Therefore, the method analyzes the electric power indexes of different levels and the prediction effect after the electric power indexes are combined with the economic statistical data layer by layer from the total indexes to the subdivision indexes to determine the optimal electric power data index, and further analyzes the optimal combination mode of the electric power data and the economic statistical data and the optimal prediction technology. In the prediction process, the invention applies the power data based on three modes (only using the current date information of the power data, only using the historical information of the power data, and using the current date and the historical information of the power data) to fully mine the value of the power data.
In the embodiment of the invention, the index of the same proportion increase rate is adopted for all the variables to eliminate the influence of seasonality, and the stability of all the variables is ensured by adopting an ADF unit root inspection technology. Regarding the processing of the stationarity of the variables, if the ADF technology indicates that a certain variable is not stationary, the variable is differentiated and the stationarity test is carried out again, and if the variable is not stationary, the differentiated time series is differentiated again.
The GDP current prediction method based on the big power data aims to solve the problems that the timeliness of the interpretation variable data is weak and valuable information cannot be fully utilized in the existing GDP prediction based on the big power data, the metrological and economics mixing technology, the big data dimension reduction technology, the prediction combination technology and the like so as to realize more timely and accurate prediction of the GDP.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (10)

1. A GDP current prediction method based on electric power big data is characterized by comprising the following steps:
step 1, constructing a plurality of GDP current prediction models based on historical sample data, wherein the historical sample data comprises economic statistical data and/or power data of a historical time period;
step 2, training a plurality of GDP current prediction models by using different prediction methods based on historical sample data to obtain a plurality of combined mean square prediction errors of the different prediction methods and the plurality of GDP current prediction models;
step 3, determining an optimal combination according to the mean square prediction errors of all combinations;
step 4, performing current prediction on the GDP by using the optimal combination and economic statistical data and/or electric power data of a preset time period;
the different prediction methods include a mixed data sampling model method, a prediction combining method, and a dynamic factor method.
2. The method of claim 1, wherein the GDP temporal prediction models include a first model, a second model, a third model, a fourth model, and a fifth model.
3. The method of claim 2, wherein the first model predicts the value of the predicted target yth period using traditional economic statistics; calculating and obtaining a first model of the predictive model by equation (1):
Figure FDA0003290754810000011
wherein the content of the first and second substances,
Figure FDA0003290754810000012
an information set of statistical data for the prediction;
Figure FDA0003290754810000013
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
hxto predict the step size, when hxWhen the current prediction is equal to 1 or 2, the information of 2 months before the current period and 1 month before the current period are respectively called for prediction.
4. The method of claim 2, wherein the second model predicts a value of the predicted target yth period using conventional economic statistics and power data; during prediction, statistical data prediction is used as a main part, and power data prediction is used as an auxiliary part; calculating and obtaining a second model of the predictive model by equation (2):
Figure FDA0003290754810000021
wherein the content of the first and second substances,
Figure FDA0003290754810000022
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure FDA0003290754810000023
representing a set of power data information used in the prediction;
Figure FDA0003290754810000024
the used ith power data index; m represents the frequency multiplication difference, epsilon, of the explanatory variable and the predicted target variableTRepresenting a prediction residual, T representing a period of a prediction target value to be predicted;
Figure FDA0003290754810000025
this indicates that prediction is performed using information that is contained in the power data but not in the statistical data.
5. The method of claim 2, wherein the third model is a prediction of the value of the tth period of the prediction target yy using power data; calculating and obtaining a third model of the prediction model by equation (3):
Figure FDA0003290754810000026
predicting the Tth period value of the prediction target y by using the power data; in the model
Figure FDA0003290754810000027
Representing a set of power data information used in the prediction;
Figure FDA0003290754810000028
Figure FDA0003290754810000031
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
when predicting step length hzEqual to 0, 1 or 2 is the current prediction, and the information of the current 3 months, the previous 2 months and the previous 1 month is used for prediction.
6. The method of claim 2, wherein the fourth model is used for predicting the Tth period value of the prediction target y by using traditional economic statistical data and power data, and the prediction is mainly based on power data prediction and is assisted by statistical data prediction; calculating and obtaining a fourth model of the predictive model by equation (4):
Figure FDA0003290754810000032
wherein the content of the first and second substances,
Figure FDA0003290754810000033
an information set of statistical data for the prediction;
Figure FDA0003290754810000034
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure FDA0003290754810000035
representing a set of power data information used in the prediction;
Figure FDA0003290754810000036
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t represents a period in which a prediction target value to be predicted is present;
Figure FDA0003290754810000037
this indicates that prediction is performed using information that is included in the statistical data but not included in the power data.
7. The method according to claim 2, wherein the fifth model is a prediction of the value of the tth period of the prediction target yy using power data; a fifth model of the prediction model is calculated and obtained by equation (5):
Figure FDA0003290754810000041
wherein the content of the first and second substances,
Figure FDA0003290754810000042
an information set of statistical data for the prediction;
Figure FDA0003290754810000043
interpreting the value of the variable at the t stage for the ith economic statistic class used for prediction;
Figure FDA0003290754810000044
represents a set of power data information used in the prediction,
Figure FDA0003290754810000045
the value of the used ith power index in the t period; m represents the frequency multiple difference of the explained variable and the predicted target variable; epsilonTRepresents the prediction residual; t denotes the period of time in which the prediction target value to be predicted is present.
8. The method of claim 1, wherein in step 2, training the GDP current prediction models by using a mixed data sampling model method based on historical sample data to obtain mean square prediction errors of various combinations of the mixed data sampling model method and the GDP current prediction models, respectively, comprises:
training N combinations formed by any one mixing data sampling model and N GDP current prediction models based on historical sample data, and obtaining mean square prediction errors of the N combinations;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model or an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
9. The method of claim 1, wherein in step 2, training the GDP current prediction models by using a prediction combination method based on historical sample data to obtain a mean square prediction error of a plurality of combinations of the prediction combination method and the GDP current prediction models, respectively, comprises:
respectively training M multiplied by N combinations formed by M mixing data sampling models and N prediction models based on historical sample data, and obtaining mean square prediction errors of the M multiplied by N combinations;
calculating the predicted values of the M multiplied by N combinations of the frequency mixing data sampling model method and the prediction model by using a prediction combination method, and calculating the mean square prediction error of the prediction combination;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model and an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
10. The method of claim 1, wherein in step 2, training the GDP current prediction models by using a dynamic factor method based on historical sample data to obtain mean square prediction errors of various combinations of the dynamic factor method and the GDP current prediction models comprises:
calculating a common factor by using the dynamic factor model;
respectively training N combinations formed by any one of the mixing data sampling models and N prediction models by taking the common factor as a new high-frequency interpretation variable, and obtaining the mean square prediction error of the N combinations;
the frequency mixing data sampling model method comprises a univariate MIDAS model, a MIDAS-AR model, an M (n) -MIDAS model or an M (n) -MIDAS-AR model;
the M (n) -MIDAS model is constructed based on a univariate MIDAS model, and the M (n) -MIDAS-AR model is constructed based on a MIDAS-AR model.
CN202111162746.9A 2021-09-30 2021-09-30 GDP (graphics device performance) current prediction method based on electric power big data Pending CN114066004A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111162746.9A CN114066004A (en) 2021-09-30 2021-09-30 GDP (graphics device performance) current prediction method based on electric power big data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111162746.9A CN114066004A (en) 2021-09-30 2021-09-30 GDP (graphics device performance) current prediction method based on electric power big data

Publications (1)

Publication Number Publication Date
CN114066004A true CN114066004A (en) 2022-02-18

Family

ID=80234035

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111162746.9A Pending CN114066004A (en) 2021-09-30 2021-09-30 GDP (graphics device performance) current prediction method based on electric power big data

Country Status (1)

Country Link
CN (1) CN114066004A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114756605A (en) * 2022-06-14 2022-07-15 之江实验室 Frequency mixing prediction method and system based on time series data

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114756605A (en) * 2022-06-14 2022-07-15 之江实验室 Frequency mixing prediction method and system based on time series data
CN114756605B (en) * 2022-06-14 2022-09-16 之江实验室 Frequency mixing prediction method and system based on time series data

Similar Documents

Publication Publication Date Title
Lin et al. Energy demand in China: Comparison of characteristics between the US and China in rapid urbanization stage
Xiao et al. A combined forecasting approach based on fuzzy soft sets
CN103258069B (en) A kind of Forecasting Methodology of steel industry electricity needs
CN112446534B (en) Construction period prediction method and device for power transmission and transformation project
CN109726865A (en) Customer charge probability density prediction technique, device and storage medium based on EMD-QRF
CN108416466A (en) Methods of electric load forecasting, the computer information processing system of complex characteristics influence
Wang et al. Application of grey model in influencing factors analysis and trend prediction of carbon emission in Shanxi Province
Bamisile et al. Comprehensive functional data analysis of China’s dynamic energy security index
CN113506121A (en) Analysis method and device for price influence factors
CN116187835A (en) Data-driven-based method and system for estimating theoretical line loss interval of transformer area
Xiang Using ARIMA‐GARCH Model to Analyze Fluctuation Law of International Oil Price
CN114066004A (en) GDP (graphics device performance) current prediction method based on electric power big data
Franta et al. Forecasting Czech GDP using mixed-frequency data models
Li Retracted: Neural network economic forecast method based on genetic algorithm
CN113011779A (en) Energy consumption price compensation method and device based on fuzzy comprehensive evaluation
CN112446551A (en) Power grid investment measuring and calculating method and device, electronic equipment and storage medium
Mo An improved ARIMA method based on hybrid dimension reduction and BP neural network
Sewata et al. Forecasting Model for Foreign Currency Exchange Rates
Lu et al. Research on time series data prediction based on clustering algorithm-A case study of Yuebao
Wei et al. Research and Development Investment Combination Forecasting Model of High‐Tech Enterprises Based on Uncertain Information
CN114091731A (en) Industrial added value current prediction method and system based on electric power big data
Wijayanti et al. Application of singular spectrum analysis method in forecasting indonesia composite data
Karunadhika et al. A Dynamic Factor Approach to Forecasting the Index of Industrial Production of Sri Lanka
CN117993959A (en) Method for predicting material price of power grid engineering equipment
Sheng et al. Analysis of time series for Malaysian currency exchange rate to the United States currency

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