Industrial transfer analysis method of support vector machine based on energy data

Abstract

The invention discloses an industrial transfer analysis method of a support vector machine based on energy data, which depends on the consistency of industrial transfer and regional energy consumption trend, extracts energy consumption data as key feature data, and further simplifies the industrial transfer analysis method. The industrial transfer analysis method is designed from two dimensions: 1, constructing an industrial geographic concentration index, and measuring the concentration or diffusion trend of key industries in a regional layout; 2, industrial output values are extracted as independent variables, a support vector machine regression model is adopted to construct a prediction model, and regional power consumption demands are predicted; and the two are combined to analyze the internal relationship between the industry transfer and the energy consumption data to assist in decision making.

Description

Technical Field

[0001] The invention relates to the technical field of energy data analysis, and in particular to an industrial transfer analysis method based on a support vector machine of energy data. Background Art

[0002] Industrial transfer is an important economic phenomenon that occurs between regions with different economic development levels. It means that under market economy conditions, some enterprises in developed regions adapt to the changes in regional comparative advantages and transfer the production of some industries to developing regions through cross-regional direct investment, thus showing the phenomenon of the industry shifting from developed regions to developing regions in terms of spatial distribution. From the perspective of my country's current economic development situation, industrial transfer has and will continue to have a far-reaching impact on my country. It is an effective way to optimize the spatial layout of productivity and form a reasonable industrial division of labor system. It is also an inevitable requirement for promoting industrial structure adjustment and accelerating the transformation of economic development mode.

[0003] Industrial transfer also brings a series of negative effects, such as industrial hollowing out, excessive competition, environmental degradation, etc. In particular, the central and western regions are often areas with relatively fragile ecological environment in my country. Taking over low-efficiency, high-pollution, and high-emission production capacity will increase energy consumption in the region and may also lead to an increase in the total energy consumption and total carbon emissions in the country, which will inevitably have a certain impact on the realization of my country's total energy consumption control targets and energy conservation and emission reduction targets.

[0004] The current quantitative analysis methods for industrial transfer have the following shortcomings:

[0005] (1) There are many factors affecting industrial transfer, such as resource costs, regional economy, transportation facilities, market conditions, and policy guidance. It is necessary to extract key characteristics related to specific industries for analysis. The characteristics of high-tech industries and traditional industries are very different, and the secondary industry is closely related to the primary industry. These factors lead to the need for more data support in the existing industrial transfer analysis methods, which are not flexible enough.

[0006] (2) The analysis of industrial transfer requires a large amount of data support, including economic data, industrial policies, corporate behavior and other information. Economic data and corporate behavior are internal information of enterprises, which are difficult to collect and costly, and may not be obtained in the end. If these key data cannot be obtained, or the obtained data is not comprehensive enough, it will directly affect the accuracy and reliability of data analysis.

[0007] (3) In most analytical methods, the spatial Gini coefficient is used to measure the degree of industrial spatial agglomeration, but this method has certain distortions; for example, when there is only one large enterprise in a certain place, this industry may have a higher Gini coefficient, but in fact no industrial agglomeration phenomenon has been formed;

[0008] (4) The existing industrial transfer analysis methods are relatively complex. They require professional knowledge and technical support from multiple dimensions such as economy, society, and environment, and are quite complex and difficult. Summary of the invention

[0009] The purpose of the present invention is to solve the shortcomings of the prior art and propose an industrial transfer analysis method based on a support vector machine of energy data. The method relies on the consistency of industrial transfer and regional energy consumption trends, extracts energy consumption data as key feature data, and further simplifies the industrial transfer analysis method; the industrial transfer analysis method is designed from two dimensions. One is to construct an industrial geographic concentration index to measure the agglomeration or diffusion trend of key industries in the regional map; the other is to extract the industrial output value as an independent variable, and use a support vector machine regression model to construct a prediction model to predict regional electricity demand; the two are combined to analyze the intrinsic relationship between industrial transfer and energy consumption data to assist decision-making.

[0010] The present invention provides an industrial transfer analysis method based on a support vector machine of energy data, comprising the following steps:

[0011] S1 Collection and processing of data related to industrial agglomeration: Determine the scope of analytical data related to industrial agglomeration, collect and process data; sort out the industry list and enterprise user list in the region, and divide the region and industry to which the enterprise belongs according to the enterprise attributes; process the collected user energy consumption data according to the regional and industrial dimensions, and generate a wide table of detailed energy consumption data for enterprises, industries, and regions;

[0012] S2 Industrial Geographic Concentration Index Algorithm Construction: The industrial geographic concentration index is constructed by integrating the spatial Gini coefficient and the Herfindahl index to measure the agglomeration or diffusion trend of key industries in the regional map; when the agglomeration index is less than 0.02, it indicates that the industrial agglomeration degree is low; when the index is between 0.02 and 0.05, the industrial agglomeration is moderate; when the index is greater than 0.05, the industrial agglomeration degree is high;

[0013] S3 support vector machine regression related data collection and processing: determine the scope of analysis data related to regional energy demand forecast, and collect and process data; according to the regional industrial situation, collect and sort out the corresponding industrial output value, total social electricity consumption and secondary industry electricity consumption data in the past 10 years, and perform data standardization processing;

[0014] S4 Support Vector Machine Regression Prediction Algorithm Construction: From the analysis of the collected data set, there are problems such as small data set and nonlinearity. According to the data situation, the SVMR prediction model with sparsity is selected to support the prediction of a small amount of data, and the model training data is fully covered and has strong stability. At the same time, the support vector can be used to understand the impact of data on the prediction results, which can further reflect the value of data;

[0015] S5. Decision support: Based on the above analysis results, construct an overall industrial development transfer analysis to assist decision making.

[0016] Preferably, step S1 includes the following sub-steps:

[0017] S1.1 Determine the scope of analysis data. The user range can be selected according to specific needs, such as 10kV and above high-voltage users;

[0018] S1.2 Data quality processing: Data processing of basic data such as monthly electricity consumption of users, including data cleaning and data governance, etc. This step improves the integrity and availability of data;

[0019] S1.3 Data dimension division: Using the user's electricity address, power supply unit and other information, a matching model is constructed through a nearest neighbor algorithm to calculate the user's administrative area and obtain user partition data;

[0020] S1.4 calculates the regional industry monthly data; based on the user partition data in 1.3, sorts out the user electricity classification information and monthly electricity consumption data in the corresponding partition, and calculates the monthly electricity consumption data of a certain industry in the whole region and the partition;

[0021] S1.5 calculates monthly user data; selects monthly electricity consumption data of enterprise users in the past three years, and integrates data on administrative regions, key industries, and other dimensions of high-voltage users;

[0022] S1.6 Prepare a wide table of detailed data; form a wide table of detailed data for a total of 36 months.

[0023] Preferably, in step S2, the industry geographic concentration index needs to be calculated based on the wide table of detailed data formed in S1; on the basis of the spatial Gini coefficient, the Herfindahl index is added to solve the problem of spatial Gini coefficient distortion and improve the accuracy of the data;

[0024] S2.1 The formula for the spatial Gini coefficient is:

[0025]

[0026] The value of G is between 0 and 1. If the value of G is closer to 0, the industrial distribution in the region is more balanced. If there is agglomeration, the value is not 0. If the value of G is closer to 1, the degree of industrial agglomeration is stronger.

[0027] S2.2 The Herfindahl Index formula is:

[0028]

[0029] Combining the above two formulas, we generate an agglomeration index to measure the geographic concentration of industries;

[0030] S2.3 Agglomeration index calculation formula is:

[0031]

[0032] Where: x i represents the proportion of electricity consumption of all enterprises in region i to the total electricity consumption of all regions; s i It indicates the proportion of electricity consumption of a certain industry in region i to the total electricity consumption of the industry in all regions; according to the actual data, It represents the distribution of enterprise size in the industry based on electricity consumption. represents the market share of the jth enterprise, where X represents the total electricity consumption in the region, and X j represents the electricity consumption of enterprise j, and N represents the number of enterprises in the industry. Based on actual data acquisition considerations, enterprise electricity consumption data is used instead of enterprise employment data;

[0033] When the agglomeration index is less than 0.02, it indicates that the degree of industrial agglomeration is low; when the index is between 0.02 and 0.05, the industrial agglomeration is moderate; when the index is greater than 0.05, the degree of industrial agglomeration is high.

[0034] Preferably, step S3 includes the following sub-steps:

[0035] S3.1 Data collection: According to the regional industrial situation, collect the corresponding industrial output value, total social electricity consumption and secondary industry electricity consumption data;

[0036] S3.2 Data standardization processing; The collected time series sample data has multiple dimensions, and their dimensions and orders of magnitude are different; through standardization processing, it can be ensured that the variables have the same scale and the degree of influence on electricity consumption is consistent; the data standardization process is:

[0037] First calculate the sample standard deviation, the formula is:

[0038]

[0039] Then calculate the standardized data, the formula is:

[0040]

[0041] Among them, x′ is the standardized data, x is the variable value, μ is the sample mean, and σ is the sample standard deviation.

[0042] Preferably, in step S4, the prediction is based on the output value and electricity consumption data of each industry in each region for at least the past 10 years; from the analysis of the collected data set, there are situations such as small data set and nonlinearity; according to the data situation, the SVMR prediction model with sparsity is selected; in this model, if the data point is close enough to the regression model, that is, it falls within the interval boundary of the regression model, then the data is not calculated loss, which is suitable for the prediction of a small amount of data; the model training data is fully covered and has strong stability. The support vector is used to understand the impact of data on the prediction results, which can further reflect the value of data;

[0043] Define the regression model function: f(x) = w T x+b, the interval boundary of the function is f(x)±ε; that is, the problem is formalized as:

[0044]

[0045] Among them, C is the regularization constant, γ is the ε-insensitive loss, defined as:

[0046]

[0047] Introducing the slack variable ε i :

[0048]

[0049] stf(x i )-y i ≤ε+ε i ,y i -f(x i )≤ε+δ i , ε i ≥0,δ i ≥0, i=0,1,2,...,m

[0050] Then the Lagrangian function L is:

[0051]

[0052] Find partial derivatives:

[0053]

[0054] C=α i +μ i

[0055]

[0056] The dual problem L′:

[0057]

[0058] KKT Conditions:

[0059]

[0060] The SCMR solution is of the form:

[0061]

[0062] Solution for b:

[0063]

[0064] Feature Mapping:

[0065]

[0066] Substitute f(x) into the final support vector machine regression algorithm:

[0067]

[0068] in, is the kernel function, and the nonlinear regression result can be obtained by substituting the corresponding independent variable values.

[0069] The advantages and technical effects of the present invention are:

[0070] (1) The present invention solves the problem of Gini coefficient distortion by using a support vector machine based on energy data. It constructs an industrial geographical concentration index by integrating the spatial Gini coefficient and the Herfindahl index. It analyzes the agglomeration or diffusion trend of key regional industries based on electricity consumption data. Based on the current status of regional industries, it extracts the output value of key industries in the region as an independent variable, and uses a support vector machine regression model to construct a prediction model for the total social electricity consumption and the secondary industry electricity consumption, so as to predict the future regional electricity demand.

[0071] (2) The present invention selects specific energy data related to the industry as characteristic variables, simplifies the factors affecting industrial transfer, performs characteristic analysis on the data, and supports multi-dimensional industrial concentration and transfer trend analysis;

[0072] (3) The characteristic variables selected by the present invention have the characteristics of controllable data sources, high data quality, and complete data cycles. Based on this data, the inherent relationship between industrial transfer and energy data is analyzed through prediction algorithms to provide relevant business support;

[0073] (4) On the basis of the spatial Gini coefficient, the Herfindahl index is added to solve the problem of spatial Gini coefficient distortion and improve the accuracy and reliability of analytical data; BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1The figure is a flow chart of the method of the present invention.

[0075] Figure 2 The graph of the predicted results of the total electricity consumption of the society of the present invention (100 million kWh);

[0076] Figure 3 It is the forecast result of the electricity consumption of the secondary industry of the present invention (billion kWh). DETAILED DESCRIPTION

[0077] The embodiments of the invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the invention, and cannot be understood as limiting the invention.

[0078] The present invention provides an industrial transfer analysis method based on a support vector machine of energy data, comprising the following steps:

[0079] S1 Collection and processing of data related to industrial agglomeration: Determine the scope of analytical data related to industrial agglomeration, collect and process data; sort out the industry list and enterprise user list in the region, and divide the region and industry to which the enterprise belongs according to the enterprise attributes; process the collected user energy consumption data according to the regional and industrial dimensions, and generate a wide table of detailed energy consumption data for enterprises, industries, and regions;

[0080] S2 Industrial Geographic Concentration Index Algorithm Construction: The industrial geographic concentration index is constructed by integrating the spatial Gini coefficient and the Herfindahl index to measure the agglomeration or diffusion trend of key industries in the regional map; when the agglomeration index is less than 0.02, it indicates that the industrial agglomeration degree is low; when the index is between 0.02 and 0.05, the industrial agglomeration is moderate; when the index is greater than 0.05, the industrial agglomeration degree is high;

[0081] S3 support vector machine regression related data collection and processing: determine the scope of analysis data related to regional energy demand forecast, and collect and process data; according to the regional industrial situation, collect and sort out the corresponding industrial output value, total social electricity consumption and secondary industry electricity consumption data in the past 10 years, and perform data standardization processing;

[0082] S4 Support Vector Machine Regression Prediction Algorithm Construction: From the analysis of the collected data set, there are problems such as small data set and nonlinearity. According to the data situation, the SVMR prediction model with sparsity is selected to support the prediction of a small amount of data, and the model training data is fully covered and has strong stability. At the same time, the support vector can be used to understand the impact of data on the prediction results, which can further reflect the value of data;

[0083] S5. Decision support: Based on the above analysis results, construct an overall industrial development transfer analysis to assist decision making.

[0084] Preferably, step S1 includes the following sub-steps:

[0085] S1.1 Determine the scope of analysis data. The user range can be selected according to specific needs, such as 10kV and above high-voltage users;

[0086] S1.2 Data quality processing: Data processing of basic data such as monthly electricity consumption of users, including data cleaning and data governance, etc. This step improves the integrity and availability of data;

[0087] S1.3 Data dimension division: Using the user's electricity address, power supply unit and other information, a matching model is constructed through a nearest neighbor algorithm to calculate the user's administrative area and obtain user partition data;

[0088] S1.4 calculates the regional industry monthly data; based on the user partition data in 1.3, sorts out the user electricity classification information and monthly electricity consumption data in the corresponding partition, and calculates the monthly electricity consumption data of a certain industry in the whole region and the partition;

[0089] S1.5 calculates monthly user data; selects monthly electricity consumption data of enterprise users in the past three years, and integrates data on administrative regions, key industries, and other dimensions of high-voltage users;

[0090] S1.6 Prepare a wide table of detailed data; form a wide table of detailed data for a total of 36 months.

[0091] Preferably, in step S2, the industry geographic concentration index needs to be calculated based on the wide table of detailed data formed in S1; on the basis of the spatial Gini coefficient, the Herfindahl index is added to solve the problem of spatial Gini coefficient distortion and improve the accuracy of the data;

[0092] S2.1 The formula for the spatial Gini coefficient is:

[0093]

[0094] The value of G is between 0 and 1. If the value of G is closer to 0, the industrial distribution in the region is more balanced. If there is agglomeration, the value is not 0. If the value of G is closer to 1, the degree of industrial agglomeration is stronger.

[0095] S2.2 The Herfindahl Index formula is:

[0096]

[0097] Combining the above two formulas, we generate an agglomeration index to measure the geographic concentration of industries;

[0098] S2.3 Agglomeration index calculation formula is:

[0099]

[0100] Where: x i represents the proportion of electricity consumption of all enterprises in region i to the total electricity consumption of all regions; s i It indicates the proportion of electricity consumption of a certain industry in region i to the total electricity consumption of the industry in all regions; according to the actual data, It represents the distribution of enterprise size in the industry based on electricity consumption. represents the market share of the jth enterprise, where X represents the total electricity consumption in the region, and X j represents the electricity consumption of enterprise j, and N represents the number of enterprises in the industry. Based on actual data acquisition considerations, enterprise electricity consumption data is used instead of enterprise employment data;

[0101] When the agglomeration index is less than 0.02, it indicates that the degree of industrial agglomeration is low; when the index is between 0.02 and 0.05, the industrial agglomeration is moderate; when the index is greater than 0.05, the degree of industrial agglomeration is high.

[0102] Preferably, step S3 includes the following sub-steps:

[0103] S3.1 Data collection: According to the regional industrial situation, collect the corresponding industrial output value, total social electricity consumption and secondary industry electricity consumption data;

[0104] S3.2 Data standardization processing; The collected time series sample data has multiple dimensions, and their dimensions and orders of magnitude are different; through standardization processing, it can be ensured that the variables have the same scale and the degree of influence on electricity consumption is consistent; the data standardization process is:

[0105] First calculate the sample standard deviation, the formula is:

[0106]

[0107] Then calculate the standardized data, the formula is:

[0108]

[0109] Among them, x′ is the standardized data, x is the variable value, μ is the sample mean, and σ is the sample standard deviation.

[0110] Preferably, in step S4, the prediction is based on the output value and electricity consumption data of each industry in each region for at least the past 10 years; from the analysis of the collected data set, there are situations such as small data set and nonlinearity; according to the data situation, the SVMR prediction model with sparsity is selected; in this model, if the data point is close enough to the regression model, that is, it falls within the interval boundary of the regression model, then the data is not calculated loss, which is suitable for the prediction of a small amount of data; the model training data is fully covered and has strong stability. The support vector is used to understand the impact of data on the prediction results, which can further reflect the value of data;

[0111] Define the regression model function: f(x) = w T x+b, the interval boundary of the function is f(x)±ε; that is, the problem is formalized as:

[0112]

[0113] Among them, C is the regularization constant, γ is the ε-insensitive loss, defined as:

[0114]

[0115] Introducing the slack variable ε i :

[0116]

[0117] stf(x i )-y i ≤ε+ε i ,y i -f(x i )≤ε+δ i , ε i ≥0,δ i ≥0, i=0,1,2,...,m

[0118] Then the Lagrangian function L is:

[0119]

[0120] Find partial derivatives:

[0121]

[0122] C=α i +μ i

[0123]

[0124] The dual problem L′:

[0125]

[0126] KKT Conditions:

[0127]

[0128] The SVMR solution is of the form:

[0129]

[0130] Solution for b:

[0131]

[0132] Feature Mapping:

[0133]

[0134] Substitute f(x) into the final support vector machine regression algorithm:

[0135]

[0136] in, is the kernel function, and the nonlinear regression result can be obtained by substituting the corresponding independent variable values.

[0137] In order to more clearly illustrate the specific implementation mode of the present invention, an embodiment is provided below:

[0138] like Figure 1 As shown, the present invention designs an industrial transfer analysis method from two dimensions. The first is to construct an industrial geographic concentration index to measure the agglomeration or diffusion trend of key industries in the regional map; the second is to construct a support vector machine regression prediction algorithm, extract industrial output value as an independent variable, and predict regional electricity demand; the two are combined to analyze the intrinsic relationship between industrial transfer and energy consumption data to assist decision-making;

[0139] Figure 2 and Figure 3 According to the prediction algorithm of the present invention, the power demand result of a certain area is predicted; by comparing the predicted value with the actual value, the accuracy of the prediction algorithm is verified;

[0140] In addition, in the present invention, the data wide table constructed in step S1.6 is preferably as follows:

[0141]

[0142]

[0143] In the description of this specification, reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in the embodiment or example of the invention.

[0144] Although embodiments of the invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. An industrial transfer analysis method based on support vector machine of energy data, characterized in that: The following steps are involved: S1 Collection and processing of data related to industrial agglomeration: determine the scope of analytical data related to industrial agglomeration, and collect and process data; Sort out the industry list and enterprise user list in the region, and divide the regions and industries to which the enterprises belong according to their attributes; The collected user energy consumption data is processed according to regional and industrial dimensions to generate a wide table of detailed energy consumption data for enterprises, industries and regions; S2 Industrial Geographic Concentration Index Algorithm Construction: The industrial geographic concentration index is constructed by integrating the spatial Gini coefficient and the Herfindahl index to measure the agglomeration or diffusion trend of key industries in the regional map; when the agglomeration index is less than 0.02, it indicates that the degree of industrial agglomeration is low; When the index is between 0.02 and 0.05, the industry is moderately concentrated; when the index is greater than 0.05, the industry is highly concentrated; S3 support vector machine regression related data collection and processing: determine the scope of analysis data related to regional energy demand forecast, and collect and process data; according to the regional industrial situation, collect and sort out the corresponding industrial output value, total social electricity consumption and secondary industry electricity consumption data in the past 10 years, and perform data standardization processing; S4 Support Vector Machine Regression Prediction Algorithm Construction: From the analysis of the collected data set, there are problems such as small data set and nonlinearity. According to the data situation, the SVMR prediction model with sparsity is selected to support the prediction of a small amount of data, and the model training data is fully covered and has strong stability. At the same time, the support vector can be used to understand the impact of data on the prediction results, which can further reflect the value of data; S5. Decision support: Based on the above analysis results, construct an overall industrial development transfer analysis to assist decision making.

2. The industrial transfer analysis method based on support vector machine of energy data according to claim 1 is characterized by: The step S1 comprises the following sub-steps: S1.1 Determine the scope of analysis data. The user range can be selected according to specific needs, such as 10kV and above high-voltage users; S1.2 Data quality processing: Data processing of basic data such as monthly electricity consumption of users, including data cleaning and data governance, etc. This step improves the integrity and availability of data; S1.3 Data dimension division: Using the user's electricity address, power supply unit and other information, a matching model is constructed through a nearest neighbor algorithm to calculate the user's administrative area and obtain user partition data; S1.4 Calculate regional industry monthly data; Based on the user partition data in 1.3, sort out the user electricity classification information and monthly electricity consumption data in the corresponding partition, and calculate the monthly electricity consumption data of a certain industry in the whole district and the partition; S1.5 calculates monthly user data; selects monthly electricity consumption data of enterprise users in the past three years, and integrates data on administrative regions, key industries, and other dimensions of high-voltage users; S1.6 Prepare a wide table of detailed data; form a wide table of detailed data for a total of 36 months.

3. The industrial transfer analysis method based on support vector machine of energy data according to claim 1 is characterized by: In step S2, the industry geographic concentration index needs to be calculated based on the detailed data wide table formed in S1; On the basis of the spatial Gini coefficient, the Herfindahl index is added to solve the problem of spatial Gini coefficient distortion and improve the accuracy of the data; S2.1 The formula for the spatial Gini coefficient is: The value of G is between 0 and 1. If the value of G is closer to 0, the industrial distribution in the region is more balanced. If there is agglomeration, the value is not 0. If the value of G is closer to 1, the degree of industrial agglomeration is stronger. S2.2 The Herfindahl Index formula is: Combining the above two formulas, we generate an agglomeration index to measure the geographic concentration of industries; S2.3 Agglomeration index calculation formula is: Where: x i represents the proportion of electricity consumption of all enterprises in region i to the total electricity consumption of all regions; s i It indicates the proportion of electricity consumption of a certain industry in region i to the total electricity consumption of the industry in all regions; according to the actual data, It represents the distribution of enterprise size in the industry based on electricity consumption. represents the market share of the jth enterprise, where X represents the total electricity consumption in the region, and X j represents the electricity consumption of enterprise j, and N represents the number of enterprises in the industry. Based on actual data acquisition considerations, enterprise electricity consumption data is used instead of enterprise employment data; When the agglomeration index is less than 0.02, it indicates that the degree of industrial agglomeration is low; When the index is between 0.02 and 0.05, the industry agglomeration is moderate; when the index is greater than 0.05, the industry agglomeration is high.

4. The industrial transfer analysis method based on support vector machine of energy data according to claim 1 is characterized by: The step S3 comprises the following sub-steps: S3.1 Data collection: According to the regional industrial situation, collect the corresponding industrial output value, total social electricity consumption and secondary industry electricity consumption data; S3.2 Data standardization processing; The collected time series sample data has multiple dimensions, and their dimensions and magnitudes are different. Through standardization, it can be ensured that the variables have the same scale and the degree of influence on electricity consumption is consistent. The data standardization process is: First calculate the sample standard deviation, the formula is: Then calculate the standardized data, the formula is: Among them, x′ is the standardized data, x is the variable value, μ is the sample mean, and σ is the sample standard deviation.

5. The industrial transfer analysis method based on support vector machine of energy data according to claim 1 is characterized by: In step S4, it is necessary to rely on the output value and electricity consumption data of each industry in each region for at least the past 10 years for prediction; from the analysis of the collected data set, there are situations such as small data set and nonlinearity; according to the data situation, a sparse SVMR prediction model is selected; in this model, if the data point is close enough to the regression model, that is, falls within the interval boundary of the regression model, then the data is not calculated for loss, which is suitable for the prediction of a small amount of data; The model training data is fully covered and highly stable. The support vectors are used to understand the impact of data on prediction results, which can further reflect the value of data. Define the regression model function: f(x) = w T x+b, the interval boundary of the function is f(x)±ε; that is, the problem is formalized as: Among them, C is the regularization constant, γ is the ε-insensitive loss, defined as: Introducing the slack variable ε i : s.t.f(x i )-y i ≤ε+ε i ,y i -f(x i )≤ε+δ i ,ε i ≥0,δ i ≥0,i=0,1,2,…,m Then the Lagrangian function L is: Find the partial derivative: C=a i +m i The dual problem L′: KKT Conditions: The SVMR solution is of the form: Solution for b: Feature Mapping: Substitute f(x) into the final support vector machine regression algorithm: in, is the kernel function, and the nonlinear regression result can be obtained by substituting the corresponding independent variable values.

Images