Method and system for acquiring influence factors of carbon emission

Abstract

The invention relates to the related field of carbon emission, in particular to a carbon emission influence factor acquisition method and system, and the method comprises the steps: carrying out the reduction of original carbon emission influence factor data through employing a neighborhood rough set attribute reduction algorithm, and obtaining a primary target influence factor; screening the primary target influence factors by adopting an improved principal component analysis method to obtain target influence factors; wherein the improved principal component analysis method comprises a correlation analysis step and a principal component analysis step, the correlation analysis step is used for performing correlation analysis on the primary target influence factors to obtain a target correlation matrix, and the principal component analysis step is used for performing correlation analysis on the primary target influence factors; and the principal component analysis step is used for performing principal component analysis on the target correlation matrix to obtain target influence factors. The method can solve the problems that in the prior art, carbon emission influence factors are large in similarity and redundancy, potential key factors are ignored, and recognition is inaccurate.

Description

Technical Field

[0001] The present invention relates to fields related to carbon emissions, and in particular to a method and system for acquiring factors affecting carbon emissions. Background Art

[0002] With the rapid growth of the global economy, the rapid increase in carbon emissions has brought about an impact on global climate change and the intensification of the greenhouse effect that cannot be ignored. Climate warming is changing weather patterns and destroying the normal balance of nature, and climate change is closely related to carbon emissions. Carbon emissions have led to rising temperatures, rising sea levels, damage to wetland and marine systems, and a decline in biodiversity. Therefore, strengthening environmental protection and reducing carbon emissions are urgent tasks.

[0003] The impact of carbon emission influencing factors on carbon emissions involves multiple research levels, which are related to internal and external environmental factors such as economic characteristics, energy consumption patterns, and geographical locations, and are characterized by high spatial and temporal complexity. Existing methods for identifying carbon emission influencing factors include traditional linear regression analysis and nonlinear analysis. In traditional linear regression analysis, econometrics combined with the STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) model are usually used to analyze carbon emission influencing factors; while nonlinear analysis methods include the use of the gray prediction model GM (1,1) to predict carbon emissions, and the analysis of carbon emission influencing factors of energy systems based on the LMDI (Logarithmic Mean DivisiIndex) model. However, the above methods do not eliminate the dependence on historical data or future trends during data processing, and do not fully consider other potential influencing factors, resulting in over-reliance on historical data and future trends, while ignoring potential key factors, resulting in inaccurate identification; at the same time, the similarity of input data is not considered, there is a lot of redundancy, and the amount of calculation is large. Summary of the invention

[0004] In order to solve the problem that the carbon emission influencing factors in the prior art are highly similar, redundant, ignore potential key factors, and lead to inaccurate identification, the present invention proposes a method for obtaining carbon emission influencing factors, including:

[0005] The neighborhood rough set attribute reduction algorithm is used to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors;

[0006] Using an improved principal component analysis method to screen the primary target influencing factors to obtain target influencing factors;

[0007] Among them, the improved principal component analysis method includes a correlation analysis step and a principal component analysis step. The correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, and the principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors.

[0008] Optionally, the step of using an improved principal component analysis method to screen the primary target influencing factors to obtain target influencing factors includes:

[0009] Standardizing the primary target influencing factors to obtain a standardized matrix;

[0010] Transforming the standardized matrix to obtain a correlation matrix;

[0011] Screening the correlation matrix based on correlation analysis to obtain a target correlation matrix;

[0012] The target correlation matrix is ​​subjected to principal component analysis to obtain target influencing factors.

[0013] Optionally, the conversion formula for converting the standardized matrix to obtain a correlation matrix is:

[0014]

[0015] Where R is the correlation matrix, X il and X lj They represent the lth observation value of the variable in the i-th row and j-th column in the standardized matrix Q, t is the number of existing influencing factors, and g is the total number of rows or columns in the correlation matrix.

[0016] Optionally, the step of performing standardization processing on the target influencing factors to obtain a standardized matrix includes:

[0017] The primary target influencing factors are standardized by using the z-score or maximum-minimum method to obtain a standardized matrix.

[0018] Optionally, the screening the correlation matrix based on the correlation analysis to obtain a target correlation matrix includes:

[0019] Using measurement system analysis method to perform correlation analysis on the elements in the correlation matrix;

[0020] The results of the correlation analysis are sorted, and the elements in the correlation matrix whose corresponding correlation analysis results are not within a preset range are discarded to obtain a target correlation matrix.

[0021] Optionally, performing principal component analysis on the target correlation matrix to obtain target influencing factors includes:

[0022] Obtaining the cumulative variance contribution rate of the characteristic roots of the target correlation matrix;

[0023] Compare the cumulative variance contribution rate with a preset value to obtain the number of principal components;

[0024] Sort the characteristic roots from large to small to obtain the unit eigenvector corresponding to the characteristic roots of the first principal components;

[0025] Using the unit eigenvector as a coefficient to perform a linear transformation on the target correlation matrix to obtain a principal component matrix;

[0026] The target influencing factors are obtained based on the principal component matrix.

[0027] Optionally, obtaining the target influencing factors based on the principal component matrix includes:

[0028] The influencing factors whose weight in each column of the principal component matrix is ​​greater than the preset value corresponding to the column are retained, and the retained influencing factors are screened in subsequent columns in this way to obtain the final target influencing factors.

[0029] Optionally, the use of a neighborhood rough set attribute reduction algorithm to simplify the original carbon emission influencing factor data to obtain primary target influencing factors includes:

[0030] The original carbon emission influencing factors are used as the condition attribute set of the neighborhood rough set attribute reduction algorithm, and the carbon emission intensity is used as the decision attribute set of the neighborhood rough set attribute reduction algorithm;

[0031] Randomly obtaining multiple subsets of the conditional attribute set;

[0032] The importance of each subset relative to the decision attribute set is obtained, and based on the importance, a plurality of subsets are screened to obtain a target subset, that is, a primary target influencing factor.

[0033] Optionally, the screening process includes: determining whether the subset satisfies the following two groups of conditions at the same time, and if so, the subset is a primary target influencing factor, wherein the two groups of conditions include:

[0034] Condition 1: The dependence of the decision attribute set on the subset is equal to the dependence of the decision attribute set on the condition attribute set;

[0035] Condition 2: The importance of any element in the subset relative to the decision attribute set is greater than 0.

[0036] Optionally, before simplifying the original carbon emission influencing factor data by using the neighborhood rough set attribute reduction algorithm to obtain the primary target influencing factors, the method further includes:

[0037] The z-score method was used to standardize the original carbon emission influencing factor data.

[0038] On the other hand, the present application also provides a system for obtaining carbon emission influencing factors, including:

[0039] A primary target influencing factor acquisition module is used to simplify the original carbon emission influencing factor data using a neighborhood rough set attribute reduction algorithm to obtain primary target influencing factors;

[0040] A target influencing factor acquisition module is used to screen the primary target influencing factors using an improved principal component analysis method to obtain target influencing factors;

[0041] Among them, the improved principal component analysis method includes a correlation analysis step and a principal component analysis step. The correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, and the principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors.

[0042] Optionally, the target influencing factor acquisition module is specifically used to:

[0043] Standardizing the primary target influencing factors to obtain a standardized matrix;

[0044] Transforming the standardized matrix to obtain a correlation matrix;

[0045] Screening the correlation matrix based on correlation analysis to obtain a target correlation matrix;

[0046] The target correlation matrix is ​​subjected to principal component analysis to obtain target influencing factors.

[0047] Optionally, the target influencing factor acquisition module converts the standardized matrix to obtain a correlation matrix using the conversion formula:

[0048]

[0049] Where R is the correlation matrix, X il and X lj They represent the lth observation value of the variable in the i-th row and j-th column in the standardized matrix Q, t is the number of existing influencing factors, and g is the total number of rows or columns in the correlation matrix.

[0050] Optionally, the target influencing factor acquisition module performs standardization processing on the target influencing factor to obtain a standardized matrix, including:

[0051] The primary target influencing factors are standardized by using the z-score or maximum-minimum method to obtain a standardized matrix.

[0052] Optionally, the target influencing factor acquisition module screens the correlation matrix based on correlation analysis to obtain a target correlation matrix, including:

[0053] Using measurement system analysis method to perform correlation analysis on the elements in the correlation matrix;

[0054] The results of the correlation analysis are sorted, and the elements in the correlation matrix whose corresponding correlation analysis results are not within a preset range are discarded to obtain a target correlation matrix.

[0055] Optionally, the target influencing factor acquisition module performs principal component analysis on the target correlation matrix to obtain target influencing factors, including:

[0056] Obtaining the cumulative variance contribution rate of the characteristic roots of the target correlation matrix;

[0057] Compare the cumulative variance contribution rate with a preset value to obtain the number of principal components;

[0058] Sort the characteristic roots from large to small to obtain the unit eigenvector corresponding to the characteristic roots of the first principal components;

[0059] Using the unit eigenvector as a coefficient to perform a linear transformation on the target correlation matrix to obtain a principal component matrix;

[0060] The target influencing factors are obtained based on the principal component matrix.

[0061] Optionally, the target influencing factor acquisition module obtains the target influencing factor based on the principal component matrix, including:

[0062] The influencing factors whose weight in each column of the principal component matrix is ​​greater than the preset value corresponding to the column are retained, and the retained influencing factors are screened in subsequent columns in this way to obtain the final target influencing factors.

[0063] Optionally, the primary target influencing factor acquisition module is specifically used to:

[0064] The original carbon emission influencing factors are used as the condition attribute set of the neighborhood rough set attribute reduction algorithm, and the carbon emission intensity is used as the decision attribute set of the neighborhood rough set attribute reduction algorithm;

[0065] Randomly obtaining multiple subsets of the conditional attribute set;

[0066] The importance of each subset relative to the decision attribute set is obtained, and based on the importance, a plurality of subsets are screened to obtain a target subset, that is, a primary target influencing factor.

[0067] Optionally, the process of the primary target influencing factor acquisition module screening multiple subsets includes: determining whether the subset satisfies the following two groups of conditions at the same time, and if so, the subset is a primary target influencing factor, wherein the two groups of conditions include:

[0068] Condition 1: The dependence of the decision attribute set on the subset is equal to the dependence of the decision attribute set on the condition attribute set;

[0069] Condition 2: The importance of any element in the subset relative to the decision attribute set is greater than 0.

[0070] Optionally, before the primary target influencing factor acquisition module uses the neighborhood rough set attribute reduction algorithm to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors, the method further includes:

[0071] The z-score method was used to standardize the original carbon emission influencing factor data.

[0072] Compared with the prior art, the present invention has the following beneficial effects:

[0073] The present invention provides a method and system for obtaining carbon emission influencing factors. The method comprises: using a neighborhood rough set attribute simplification algorithm to simplify original carbon emission influencing factor data to obtain primary target influencing factors; using an improved principal component analysis method to screen the primary target influencing factors to obtain target influencing factors; wherein the improved principal component analysis method comprises a correlation analysis step and a principal component analysis step, the correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, and the principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors; the neighborhood rough set attribute simplification algorithm of the present application adopts the neighborhood concept to consider potential factors, performs a first-level simplification, and the correlation analysis can analyze the correlation of factors to enhance the differences between influencing factors, realize the identification of similarity and redundancy of influencing factors, perform a second-level simplification, and further adopt principal component analysis. Through the principal component concept, the influencing factors are made independent of each other, mutual interference is eliminated, and further simplification is performed, thereby solving the problem of large similarity, much redundancy, and neglect of potential key factors in the prior art, resulting in inaccurate identification of carbon emission influencing factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A schematic diagram of the steps of the method for obtaining carbon emission influencing factors proposed by the present invention;

[0075] Figure 2 A schematic diagram of the application process of the neighborhood rough set attribute reduction method proposed by the present invention;

[0076] Figure 3A schematic diagram of the application process of the principal component analysis proposed in the present invention;

[0077] Figure 4 The present invention proposes Figure 1 Detailed step diagram of step S1;

[0078] Figure 5 The present invention proposes Figure 1 Detailed step diagram of step S2;

[0079] Figure 6 The present invention proposes Figure 5 Detailed step diagram of step S23;

[0080] Figure 7 The present invention proposes Figure 5 Detailed step diagram of step S24;

[0081] Figure 8 This is a schematic diagram of the structure of the system for acquiring carbon emission influencing factors proposed by the present invention. DETAILED DESCRIPTION

[0082] The present invention proposes a method for obtaining carbon emission influencing factors, which screens the influencing factors of carbon emissions step by step through a neighborhood rough set attribute simplification algorithm and an improved principal component analysis method, fully considering the continuity characteristics of the selected indicators. The neighborhood relationship of the neighborhood rough set attribute simplification algorithm and the correlation analysis in the improved principal component analysis method enhance the differences between the carbon emission influencing factors and improve the accuracy of screening the influencing factors.

[0083] Embodiment 1:

[0084] The research objects of this embodiment involve multiple systems such as industry, energy, and economy. The relevant influencing factors of industry, economy, and energy from 2017 to 2021 are collected. Based on the similarity and redundancy of the influencing factors, the neighborhood rough set reduction algorithm, correlation analysis method, and principal component analysis method are used to gradually reduce the dimension of the influencing factors, so that the results are more in line with the actual situation of the influencing factors of carbon emissions. Figure 1 As shown, the method for obtaining the carbon emission influencing factors of this application is as follows: Figure 1 As shown, the process includes the following steps S1 and S2.

[0085] S1: The neighborhood rough set attribute reduction algorithm is used to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors.

[0086] In a further preferred embodiment, Figure 2 As shown, before the neighborhood rough set attribute simplification algorithm is used to simplify the original carbon emission influencing factor data, the following steps are also included: standardizing the original carbon emission influencing factor data.

[0087] The present invention preferably uses the z-score method to standardize data. The influencing factors and corresponding data collected by the three systems of industry, economy and energy are collected, and x1, x2, ..., x n To express.

[0088] Use the z-score method to analyze the sequence x1,x2,...,x n Transformation is performed, and the transformation formula is as follows:

[0089]

[0090] in, is the average value of the data of n influencing factors, x i is the i-th influencing factor, X i is the standardized i-th influencing factor, n is the total number of influencing factors, s is the variance,

[0091] Then the new sequence X1,X2,...,X n The mean is 0, the variance is 1, and it is dimensionless. In this embodiment, the original carbon emission influencing factors of the three subsystems of industry, economy, and energy are preliminarily determined by literature reading, etc., and the relevant data from 2017 to 2021 are collected. After standardized processing in the above manner, the data are shown in Table 1 below.

[0092] Table 1: Standardized results of factors affecting raw carbon emissions

[0093]

[0094]

[0095]

[0096]

[0097] In a further preferred solution, the neighborhood rough set attribute reduction algorithm is used to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors, such as Figure 4 As shown, the following steps S11 to S13 are included:

[0098] S11: taking the original carbon emission influencing factors as the condition attribute set of the neighborhood rough set attribute reduction algorithm, and taking the carbon emission intensity as the decision attribute set of the neighborhood rough set attribute reduction algorithm;

[0099] S12: randomly acquiring multiple subsets of the conditional attribute set;

[0100] S13: Obtain the importance of each subset relative to the decision attribute set, and screen multiple subsets based on the importance to obtain a target subset, that is, a primary target influencing factor.

[0101] Step S11, see Figure 2 , including taking the original carbon emission influencing factors as the condition attribute set C, taking the carbon emission intensity as the decision attribute set D, taking the data corresponding to the original carbon emission influencing factors as the object set U, constructing a decision system <U,C∪D>; obtaining a distance function; calculating the neighborhood of the decision system based on the distance function; step S13 includes obtaining an upper approximation on a subset B of the condition attributes based on the neighborhood And the lower approximation N(D); based on the lower approximation, the positive domain value POS is obtained B (D); Obtaining neighborhood dependence γ based on the positive domain value B (D) and the importance Sig(c,B,D) of any attribute c in the subset B of conditional attributes relative to the decision attribute; determine whether the importance Sig(c,B,D) is greater than 0, if so, B is the primary target influencing factor.

[0102] The specific steps of step S1 are as follows:

[0103] In the neighborhood rough set attribute reduction algorithm, let <U,C∪D> be a neighborhood decision system, where U is the object set, that is:

[0104] U={x1,x2,...,x n} (2)

[0105] All data are continuous, C∪D is an attribute set, where C is a conditional attribute set, which corresponds to the original carbon emission influencing factors in this application; D is a decision attribute set, which corresponds to the carbon emission intensity in this application, and a subset B of the conditional attribute set is randomly obtained.

[0106] Given any x i ∈U and Define sample x i The neighborhood on subset B is:

[0107] δ B (x i )={x j ∈U|Δ B (x i ,x j )≤δ} (3)

[0108] Among them, δ is the neighborhood radius, Δ B (.) is the distance function on subset B. For a subset B of the M-dimensional conditional attribute set, Δ B (.) is usually expressed using the P norm as:

[0109]

[0110] Among them, P can be 1, 2, ∞, through different distance functions Δ B (.), can realize attribute reduction on discrete space, continuous space, and mixed space. This application is continuous control, so P = 2, M is the total dimension of the conditional attribute subset B, and k is the kth dimension of the conditional attribute subset B. In addition, when δ = 0, the neighborhood rough set attribute reduction model degenerates into a classical rough set model, so δ ≠ 0.

[0111] For a given X={X1,X2,...X r}∈U / D,X i ∈U, X is any set of influencing factors, X r is the rth item in any influencing factor set, with a neighborhood relationship of N and a neighborhood of δ B (X i ), then the upper approximation and lower approximation of the decision attribute set D on N with respect to B are:

[0112] Upper approximation:

[0113] The lower approximation is:

[0114] in:

[0115]

[0116] The lower approximation of D with respect to B is usually called the positive domain, denoted as POS B (D).

[0117] The domain dependency of decision attribute set D on B can be defined through the positive domain as:

[0118]

[0119] From the above formula (9), we can see that the dependency is monotonic, so the importance calculation formula of any attribute c in B relative to D is:

[0120] Sig(c,B,D)=γ B (D)-γ B-{c} (D) (10)

[0121] γ B-{c} (D) is the domain dependence of the decision attribute set D after removing attribute c from B.

[0122] In a further preferred solution, the importance of each subset relative to the decision attribute set is obtained, and multiple subsets are screened based on the importance to obtain a target subset. The screening process includes: determining whether the subset satisfies the following two groups of conditions at the same time. If so, the subset is a primary target influencing factor, wherein the two groups of conditions include:

[0123] Condition 1: The dependency of the decision attribute set on the subset of the condition attribute set is equal to the dependency of the decision attribute set on the condition attribute set;

[0124] Condition 2: The importance of any element in the subset of the condition attribute set relative to the decision attribute set is greater than 0.

[0125] Based on the above conditions, in this embodiment, if subset B meets the following conditions:

[0126] (1)γ B (D) = γ C (D), that is, Pos B (D)=Pos C (D), it means that the classification capabilities of the conditional attribute subsets B and C are the same; γ C (D) is the domain dependence of D on C.

[0127] (2) γ B (D)>γ B-{c} (D), that is, Sig(c,B,D)>0, which means that c is important relative to D.

[0128] If the subset B of the conditional attributes satisfies the above two conditions, then B is a simplification of C. Based on the above method, the primary target influencing factors are obtained. In this embodiment, the primary target influencing factors are shown in Table 2 below.

[0129] Table 2: Primary target influencing factors

[0130]

[0131]

[0132] S2: Using an improved principal component analysis method to screen the primary target influencing factors to obtain target influencing factors.

[0133] Wherein, the improved principal component analysis method includes a correlation analysis step and a principal component analysis step.

[0134] The correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, such as Figure 3As shown, the correlation analysis step includes performing correlation analysis on the primary target influencing factors to obtain a target correlation matrix.

[0135] The principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors, specifically, Figure 3 As shown, it includes calculating the variance contribution rate of the characteristic root of each indicator based on the target correlation matrix, obtaining the principal component based on the variance contribution rate, judging whether the cumulative variance contribution rate of the principal component is greater than the preset value, and if so, obtaining the principal component matrix, and obtaining the target influencing factor based on the principal component matrix; if not, reselecting the principal component, and rejudging the relationship between the cumulative variance contribution rate of the principal component and the preset value, until the cumulative variance contribution rate of the principal component is greater than the preset value.

[0136] In a further preferred embodiment, Figure 5 As shown, step S2 includes the following steps S21 to S24:

[0137] S21: performing standardization processing on the primary target influencing factors to obtain a standardized matrix;

[0138] S22: transforming the standardized matrix to obtain a correlation matrix;

[0139] S23: Screening the correlation matrix based on correlation analysis to obtain a target correlation matrix;

[0140] S24: Performing principal component analysis on the target correlation matrix to obtain target influencing factors.

[0141] In a further preferred solution, in step S21, the primary target influencing factors can be standardized by using a z-score or a maximum-minimum method to obtain a standardized matrix.

[0142] In a further preferred solution, in step S22, the conversion formula for converting the standardized matrix to obtain the correlation matrix is:

[0143]

[0144] Where R is the correlation matrix, X il and X lj They represent the lth observation value of the variable in the i-th row and j-th column in the standardized matrix Q, t is the number of existing influencing factor variables, and g is the number of rows or columns in the correlation matrix.

[0145] In a further preferred solution, in step S23, the correlation matrix is ​​screened based on the correlation analysis to obtain a target correlation matrix, such as Figure 6 As shown, it includes step S231 and step S232:

[0146] S231: Perform correlation analysis on the elements in the correlation matrix using a measurement system analysis method.

[0147] Measurement System Analysis (MSA) is used to characterize the correlation between a certain item and other items. Therefore, in this application, MSA is used to characterize the correlation between a certain element in the correlation matrix and other elements. The expression of MSA is:

[0148] MSA=A / (A+B) (12)

[0149] Wherein, A is the sum of squares of correlation coefficients between a certain element and other elements, and B is the sum of squares of partial correlation coefficients between the element and other elements. The process of solving the correlation coefficient and the partial correlation coefficient is prior art and will not be repeated here. In this embodiment, the results of the correlation analysis are shown in Table 3 below.

[0150] Table 3: Results of correlation analysis

[0151]

[0152]

[0153] S232: Sort the results of the correlation analysis, and discard the elements in the correlation matrix whose corresponding correlation analysis results are not within a preset range, so as to obtain a target correlation matrix.

[0154] The results of the correlation analysis are sorted, and the elements whose corresponding correlation analysis results are not within a preset range are discarded in the correlation matrix. The discarding method can be to fill the discarded elements in the correlation matrix with 0. For example, elements with MSA values ​​less than 0.2 indicate that they are of little help in information concentration, and elements with MSA values ​​greater than 0.8 indicate that the information overlaps too much with the remaining elements. Therefore, elements with MSA values ​​less than 0.2 and greater than 0.8 are discarded to obtain the target correlation matrix R*.

[0155] S24: Performing principal component analysis (PCA) on the target correlation matrix to obtain target influencing factors.

[0156] In a further preferred embodiment, Figure 7 As shown, step S24 includes the following steps S241 to S245.

[0157] S241: Obtain the cumulative variance contribution rate of the characteristic roots of the target correlation matrix;

[0158] The characteristic equation of the target correlation matrix R* is solved to obtain the m characteristic roots and the corresponding m unit eigenvectors of the target correlation matrix R*.

[0159] The characteristic equation of the target correlation matrix R* is:

[0160] |R*-λI|=0 (13)

[0161] Among them, λ is the eigenvalue vector and I is the unit vector.

[0162] The m eigenvalues ​​of the target correlation matrix R* are:

[0163] λ1≥λ2≥...≥λ m (14)

[0164] For characteristic root λ i The variance contribution rate η of the first k characteristic roots i Accumulate to obtain the cumulative variance contribution rate

[0165] S242: Compare the cumulative variance contribution rate with a preset value to obtain the number of principal components.

[0166] When the cumulative variance contribution rate is just greater than or equal to the preset value, the corresponding cumulative number of characteristic roots is the number of principal components k. Therefore, k≤m. In this embodiment, the number of principal components is 4. The characteristic roots and cumulative contribution rates corresponding to the principal components are shown in Table 4 below.

[0167] Table 4: Principal components, characteristic roots and contribution rates

[0168] principal component Eigenvalue Contribution rate Cumulative contribution rate The first principal component 27.955 71.679 71.679 The second principal component 6.317 16.197 87.876 The third principal component 3.356 8.605 96.482 The fourth principal component 1.372 3.518 100.000

[0169] S243: Sort the characteristic roots from large to small to obtain the unit eigenvectors corresponding to the characteristic roots of the first principal components.

[0170] Get the unit eigenvector a corresponding to the first k eigenroots f :

[0171] a f =(a 1f ,a 2f ,…,a mf ) T ,f=1,2,…,k (15)

[0172] Among them, a mf is the element of the mth row of the fth eigenvector.

[0173] S244: Use the unit eigenvector as a coefficient to perform a linear transformation on the target correlation matrix to obtain a principal component matrix.

[0174]

[0175] Wherein, Y is the principal component matrix. In this embodiment, the principal component matrix is ​​shown in Table 5 below:

[0176] Table 5: Principal component matrix

[0177]

[0178]

[0179] Then, the i-th principal component value of the j-th column of the target correlation matrix is:

[0180]

[0181] Y ij Corresponding to the values ​​in Table 5, that is, the corresponding weights, (X 1j ,X 2j ,…,X mj ) T is the j-th column of the target correlation matrix, e is the total number of columns of the target correlation matrix, and m is the number of eigenvalues ​​of the target correlation matrix R*.

[0182] S245: Obtain target influencing factors based on the principal component matrix.

[0183] In a further preferred solution, the step S245 is specifically as follows:

[0184] The influencing factors whose weight in each column of the principal component matrix is ​​greater than the preset value corresponding to the column are retained, and the retained influencing factors are screened in subsequent columns in this way to obtain the final target influencing factors.

[0185] The 4 columns in Table 5 correspond to 4 preset values, namely the first preset value, the second preset value, the third preset value and the fourth preset value. First, filter from the first column on the left, and retain the data with an absolute value greater than or equal to the first preset value. Continue to filter the second column, the third column and the fourth column in turn for the retained data to finally obtain the target influencing factors.

[0186] Embodiment 2:

[0187] The present invention based on the same inventive concept also provides a system for obtaining carbon emission influencing factors, such as Figure 8 As shown, it includes a primary target influencing factor acquisition module and a target influencing factor acquisition module, wherein:

[0188] A primary target influencing factor acquisition module is used to simplify the original carbon emission influencing factor data using a neighborhood rough set attribute reduction algorithm to obtain primary target influencing factors;

[0189] A target influencing factor acquisition module is used to screen the primary target influencing factors using an improved principal component analysis method to obtain target influencing factors;

[0190] Among them, the improved principal component analysis method includes a correlation analysis step and a principal component analysis step. The correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, and the principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors.

[0191] In a further preferred solution, the target influencing factor acquisition module is specifically used to:

[0192] Standardizing the primary target influencing factors to obtain a standardized matrix;

[0193] Transforming the standardized matrix to obtain a correlation matrix;

[0194] Screening the correlation matrix based on correlation analysis to obtain a target correlation matrix;

[0195] The target correlation matrix is ​​subjected to principal component analysis to obtain target influencing factors.

[0196] In a further preferred solution, the target influencing factor acquisition module converts the standardized matrix to obtain a correlation matrix using the conversion formula:

[0197]

[0198] Where R is the correlation matrix, X il and X lj They represent the lth observation value of the variable in the i-th row and j-th column in the standardized matrix Q, t is the number of existing influencing factors, and g is the total number of rows or columns in the correlation matrix.

[0199] In a further preferred solution, the target influencing factor acquisition module performs standardization processing on the target influencing factor to obtain a standardized matrix, including:

[0200] The primary target influencing factors are standardized by using the z-score or maximum-minimum method to obtain a standardized matrix.

[0201] In a further preferred solution, the target influencing factor acquisition module screens the correlation matrix based on correlation analysis to obtain a target correlation matrix, including:

[0202] Using measurement system analysis method to perform correlation analysis on the elements in the correlation matrix;

[0203] The results of the correlation analysis are sorted, and the elements in the correlation matrix whose corresponding correlation analysis results are not within a preset range are discarded to obtain a target correlation matrix.

[0204] In a further preferred solution, the target influencing factor acquisition module performs principal component analysis on the target correlation matrix to obtain the target influencing factors, including:

[0205] Obtaining the cumulative variance contribution rate of the characteristic roots of the target correlation matrix;

[0206] Compare the cumulative variance contribution rate with a preset value to obtain the number of principal components;

[0207] Sort the characteristic roots from large to small to obtain the unit eigenvector corresponding to the characteristic roots of the first principal components;

[0208] Using the unit eigenvector as a coefficient to perform a linear transformation on the target correlation matrix to obtain a principal component matrix;

[0209] The target influencing factors are obtained based on the principal component matrix.

[0210] In a further preferred solution, the target influencing factor acquisition module obtains the target influencing factor based on the principal component matrix, including:

[0211] The influencing factors whose weight in each column of the principal component matrix is ​​greater than the preset value corresponding to the column are retained, and the retained influencing factors are screened in subsequent columns in this way to obtain the final target influencing factors.

[0212] In a further preferred solution, the primary target influencing factor acquisition module is specifically used to:

[0213] The original carbon emission influencing factors are used as the condition attribute set of the neighborhood rough set attribute reduction algorithm, and the carbon emission intensity is used as the decision attribute set of the neighborhood rough set attribute reduction algorithm;

[0214] Randomly obtaining multiple subsets of the conditional attribute set;

[0215] The importance of each subset relative to the decision attribute set is obtained, and based on the importance, a plurality of subsets are screened to obtain a target subset, that is, a primary target influencing factor.

[0216] In a further preferred solution, the process of screening multiple subsets by the primary target influencing factor acquisition module includes: determining whether the subset satisfies the following two groups of conditions at the same time, and if so, the subset is a primary target influencing factor, wherein the two groups of conditions include:

[0217] Condition 1: The dependence of the decision attribute set on the subset is equal to the dependence of the decision attribute set on the condition attribute set;

[0218] Condition 2: The importance of any element in the subset relative to the decision attribute set is greater than 0.

[0219] In a further preferred solution, before the primary target influencing factor acquisition module uses the neighborhood rough set attribute reduction algorithm to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors, the method further includes:

[0220] The z-score method was used to standardize the original carbon emission influencing factor data.

[0221] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

Claims

1. A method for obtaining carbon emission influencing factors, characterized in that: include: The neighborhood rough set attribute reduction algorithm is used to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors; Using an improved principal component analysis method to screen the primary target influencing factors to obtain target influencing factors; Among them, the improved principal component analysis method includes a correlation analysis step and a principal component analysis step. The correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, and the principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors.

2. The method for obtaining carbon emission influencing factors according to claim 1, characterized in that: The improved principal component analysis method is used to screen the primary target influencing factors to obtain the target influencing factors, including: Standardizing the primary target influencing factors to obtain a standardized matrix; Transforming the standardized matrix to obtain a correlation matrix; Screening the correlation matrix based on correlation analysis to obtain a target correlation matrix; The target correlation matrix is ​​subjected to principal component analysis to obtain target influencing factors.

3. The method for obtaining carbon emission influencing factors according to claim 2, characterized in that: The conversion formula for converting the standardized matrix to obtain the correlation matrix is: Where R is the correlation matrix, X il and X lj They represent the lth observation value of the variable in the i-th row and j-th column in the standardized matrix Q, t is the number of existing influencing factors, and g is the total number of rows or columns in the correlation matrix.

4. The method for obtaining carbon emission influencing factors according to claim 2, characterized in that: The step of screening the correlation matrix based on the correlation analysis to obtain a target correlation matrix includes: Using measurement system analysis method to perform correlation analysis on the elements in the correlation matrix; The results of the correlation analysis are sorted, and the elements in the correlation matrix whose corresponding correlation analysis results are not within a preset range are discarded to obtain a target correlation matrix.

5. The method for obtaining carbon emission influencing factors according to claim 2, characterized in that: The performing principal component analysis on the target correlation matrix to obtain target influencing factors includes: Obtaining the cumulative variance contribution rate of the characteristic roots of the target correlation matrix; Compare the cumulative variance contribution rate with a preset value to obtain the number of principal components; Sort the characteristic roots from large to small to obtain the unit eigenvector corresponding to the characteristic roots of the first principal components; Using the unit eigenvector as a coefficient to perform a linear transformation on the target correlation matrix to obtain a principal component matrix; The target influencing factors are obtained based on the principal component matrix.

6. The method for obtaining carbon emission influencing factors according to claim 5, characterized in that: The obtaining of target influencing factors based on the principal component matrix includes: The influencing factors whose weight in each column of the principal component matrix is ​​greater than the preset value corresponding to the column are retained, and the retained influencing factors are screened in subsequent columns in this way to obtain the final target influencing factors.

7. The method for obtaining carbon emission influencing factors according to claim 1, characterized in that: The neighborhood rough set attribute reduction algorithm is used to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors, including: The original carbon emission influencing factors are used as the condition attribute set of the neighborhood rough set attribute reduction algorithm, and the carbon emission intensity is used as the decision attribute set of the neighborhood rough set attribute reduction algorithm; Randomly obtaining multiple subsets of the conditional attribute set; The importance of each subset relative to the decision attribute set is obtained, and based on the importance, a plurality of subsets are screened to obtain a target subset, that is, a primary target influencing factor.

8. The method for obtaining carbon emission influencing factors according to claim 7, characterized in that: The screening process includes: determining whether the subset satisfies the following two groups of conditions at the same time, and if so, the subset is a primary target influencing factor, wherein the two groups of conditions include: Condition 1: The dependence of the decision attribute set on the subset is equal to the dependence of the decision attribute set on the condition attribute set; Condition 2: The importance of any element in the subset relative to the decision attribute set is greater than 0.

9. The method for obtaining carbon emission influencing factors according to any one of claims 1, 7 and 8, characterized in that: Before using the neighborhood rough set attribute simplification algorithm to simplify the original carbon emission influencing factor data to obtain the primary target influencing factors, the method further includes: The z-score method was used to standardize the original carbon emission influencing factor data.

10. A system for obtaining carbon emission influencing factors, characterized in that: include: A primary target influencing factor acquisition module is used to simplify the original carbon emission influencing factor data using a neighborhood rough set attribute reduction algorithm to obtain primary target influencing factors; A target influencing factor acquisition module is used to screen the primary target influencing factors using an improved principal component analysis method to obtain target influencing factors; Among them, the improved principal component analysis method includes a correlation analysis step and a principal component analysis step. The correlation analysis step is used to perform correlation analysis on the primary target influencing factors to obtain a target correlation matrix, and the principal component analysis step is used to perform principal component analysis on the target correlation matrix to obtain target influencing factors.

Images