Multi-variable evaluation method

Abstract

The invention relates to a multi-variable evaluation method. The method comprises the following steps: 1) performing correlation analysis on a plurality of indexes of a to-be-evaluated object, and removing indexes without significant correlation to obtain retention indexes of the to-be-evaluated object; 2) if an index numerical value with a negative change trend exists in the retention indexes, performing reciprocal valuing on the index numerical value to obtain a positive same-trend numerical value of the to-be-evaluated object index; 3) converting a forward same-trend numerical value of the to-be-evaluated object index into an index conversion numerical value in a range of [0, 1]; 4) calculating the average of the sum of the deviations of the n index conversion values of the to-be-evaluated objects and the minimum values of the corresponding index conversion values in all the to-be-evaluated objects to obtain evaluation values for evaluating the to-be-evaluated objects; and 5) the evaluation value of the to-be-evaluated object is compared with 1, and the closer the evaluation value is to 1, the better the comprehensive performance of the to-be-evaluated object is.

Description

Technical Field

[0001] This invention relates to an evaluation method, and more particularly to a multivariate evaluation method. Background Technology

[0002] Multivariate comprehensive evaluation method refers to the method of evaluating multiple objects using multiple indicators. It is also called multivariate comprehensive evaluation method (or simply comprehensive evaluation method). Essentially, it transforms multiple indicators into a single indicator that reflects the overall situation. It is generally used in social sciences and natural sciences, such as evaluating the comprehensive strength of different countries, the economic benefits of enterprises, and the growth status of plants.

[0003] A typical modern comprehensive evaluation method is the membership function method, which first performs principal component analysis on multiple indicators, then calculates the membership function values ​​and weights of each indicator to obtain the comprehensive evaluation value of each object to be evaluated. However, the meaning of principal component factors is generally difficult to grasp, and the evaluation process and steps of the membership function method are cumbersome, complex, and difficult to understand. Therefore, it is necessary to develop a simple, intuitive, practical, and applicable multivariate evaluation method. Summary of the Invention

[0004] A multivariate evaluation method includes the following steps:

[0005] 1) Perform correlation analysis on multiple indicators of the object to be evaluated, remove the indicators of the object to be evaluated that have no significant correlation with other indicators, and retain the indicators of the object to be evaluated that have significant correlation, thus obtaining the retained indicators of the object to be evaluated.

[0006] 2) If the changing trends of all the indicators in the retained indicators are positive, then there is no need to take the reciprocal value of any of the retained indicators; if there are indicators in the retained indicators with a negative changing trend, then the reciprocal value of the indicator is taken; after the positive trending transformation in this step, the positive trending value of the indicator of the object to be evaluated is obtained.

[0007] 3) Convert the positive trend values ​​of the indicators of the object to be evaluated into indicator conversion values ​​in the range of [0,1].

[0008] 4) Determine the number n of the positive trend indicators of the object to be evaluated, and then calculate the average of the sum of the deviations of the n indicator conversion values ​​of the object to be evaluated from the minimum value of the corresponding indicator conversion values ​​of all objects to be evaluated, so as to obtain the evaluation value used to evaluate the object to be evaluated.

[0009] 5) Compare the evaluation value of the object to be evaluated with 1. The closer the evaluation value is to 1, the better the overall performance of the object to be evaluated.

[0010] In one specific implementation, in step 1), the correlation coefficient is used to determine the correlation between any two indicators among the multiple indicators of the object to be evaluated, and the correlation coefficient is calculated based on formula 1):

[0011]

[0012] Where r is the correlation coefficient, i or i' is the indicator, k is the object to be evaluated, and X ik Let i be the positive trend-normalized value of index i for the object k to be evaluated. X is the average of the positively trending values ​​of indicator i for all objects to be evaluated. i'k Let i' be the positive trend-based value of the index i' of the object to be evaluated k. The average of the positively trending values ​​of indicator i' of all objects to be evaluated.

[0013] In one specific implementation, in step 3), the index conversion value is calculated based on formula 2):

[0014]

[0015] Among them, u ik For X ik The converted value of the indicator; X min(i) For X among all objects to be evaluated i The minimum value of X max(i) For X among all objects to be evaluated i The maximum value.

[0016] In one specific implementation, in step 4), the evaluation value is calculated based on formula 3):

[0017]

[0018] Among them, Z k u is the comprehensive evaluation value of the object to be evaluated. ik For X ik The converted value of the indicator, u min(i) For the object to be evaluated, u i The minimum value of , where n is the number of positive homotrend indicators of the object to be evaluated.

[0019] In one specific implementation, it also includes step 6), based on Z in step 4). k Perform clustering or sorting.

[0020] In one specific implementation, when there are N objects to be evaluated, the K-means clustering method is used to cluster the Z-axis of the objects to be evaluated. k Perform clustering and set the number of clusters L, where L < N.

[0021] The beneficial effects of this invention are as follows: This invention provides a multivariate evaluation method. First, it detects the correlation between multiple indicators of the object to be evaluated, eliminating indicators that do not have a significant correlation with other indicators, while ensuring the positive co-trend of changes in multiple indicators. Second, it transforms the positively co-trending indicators so that their values ​​are between 0 and 1. Third, it calculates the average of the deviations of all transformed values ​​of the object to be evaluated from its minimum value, using this as its comprehensive evaluation value. Finally, it clusters or sorts the comprehensive evaluation values. The steps of this method are logically coherent and progressive, with intuitive and clear meanings. The calculations are simple and easy to interpret, and the evaluation results are largely consistent with those of the membership function method, demonstrating good practicality and applicability. Detailed Implementation

[0022] The present invention will be further described in detail below through preferred embodiments, but these embodiments do not constitute a limitation thereof.

[0023] Example 1

[0024] A comprehensive evaluation was conducted on the academic performance of 55 students in a class of the second year of junior high school in Cangzhou City during the first semester.

[0025] The final exam scores for 55 students in Chinese, Mathematics, English, Physics, Biology, Politics, History, and Geography are detailed in Table 1.

[0026] Table 1

[0027]

[0028] (1) Correlation analysis: The correlation coefficient formula (1) is used to determine the pairwise correlation between the academic performance of all students in each course. X ik or X i'k The variables or observations represent academic performance, where i or i' is the subject code, i or i' = 1 for Chinese, i or i' = 2 for Mathematics, i or i' = 3 for English, i or i' = 4 for Physics, i or i' = 5 for Biology, i or i' = 6 for Politics, i or i' = 7 for History, and i or i' = 8 for Geography; k is the student code, a number from 1 to 55. Taking the correlation between the Chinese and Mathematics academic performance of all students in the class as an example, the correlation coefficient between Chinese and Mathematics academic performance is calculated based on formula (1), where r is the correlation coefficient. Let i be the average score of all students in the class for subject i. Let i' be the average score of all students in the class for subject i. When i = 1 and k = 1, the first student's Chinese score is X. 11 When i' = 2 and k = 1, the first student's math score is X. 21 , This represents the average Chinese language score of all students in the class. This is the average math score of all students in the class.

[0029]

[0030] The correlation analysis results of the students' academic performance in all subjects are shown in Table 2.

[0031] Table 2

[0032] Correlation coefficient Language math English physics biology politics history geography Language 1 math 0.6982** 1 English 0.7900** 0.7343** 1 physics 0.6983** 0.8337** 0.7276** 1 biology 0.7605** 0.6695** 0.5722** 0.7052** 1 politics 0.8326** 0.6555** 0.8629** 0.6502** 0.6376** 1 history 0.6332** 0.6832** 0.5677** 0.6940** 0.6807** 0.6001** 1 geography 0.6966** 0.6772** 0.5961** 0.6787** 0.7976** 0.6466** 0.7169** 1

[0033] Note: ** indicates that the difference is highly significant at the 0.01 level.

[0034] Table 2 shows that there are highly significant positive correlations among the scores of the students in this class across all eight subjects, including Chinese, Mathematics, and English, and all correlations exhibit a positive trend. Therefore, it is not necessary to take the reciprocal. Let the student's academic performance X... ik To conduct a comprehensive evaluation of the students' academic performance in this class.

[0035] (2) Calculate the converted values ​​based on formula (2).

[0036]

[0037] Among them, u ik For X ik The converted value; X min(i) For 55 X i The minimum value of X max(i) For 55 X i The maximum value in; similarly, i ranges from 1 to 8, and k ranges from 1 to 55.

[0038] The converted numerical results are shown in Table 3.

[0039] (3) Calculate the comprehensive evaluation value based on formula (3).

[0040]

[0041] Table 3

[0042]

[0043] Among them, Z k u represents the comprehensive evaluation score for each student. ik For X ikThe converted value, u min(i) For 55 u i The minimum value in the range is n, where n is the total number of i, which is 8 in this case. The results are shown in Table 4.

[0044] (4) Classification by level: Z k Cluster analysis was performed on the scores to classify the students into different levels. The K-means clustering method was used to assign each student's evaluation score Z to the class. k Perform clustering, setting the number of clusters to 5, i.e., Z0 k The scores are divided into five categories, corresponding to academic performance levels from highest to lowest: A (Excellent), B (Good), C (Average), D (Poor), and E (Very Poor), as shown in Table 4. Table 4 also lists each student's total score and their comprehensive evaluation score Z. k The correlation coefficient with its total score is 0.9991, indicating a highly significant positive correlation.

[0045] Table 4

[0046]

[0047] Comparative Example 1

[0048] The membership function method is used for comprehensive evaluation. The specific calculation method and process are as follows.

[0049] (1) Principal Component Analysis: Principal component analysis was conducted on students' scores in eight subjects, including Chinese, Mathematics, and English. The principal component factors obtained from the principal component analysis were used as a comprehensive indicator to evaluate students' academic performance. Two independent principal components were selected as the comprehensive indicator of students' academic performance. The `princomp()` function in R was used to conduct principal component analysis on the scores of 55 students in eight subjects to obtain the eigenvalues ​​and contribution rates (P) of each principal component. j (j = 1 or 2), see Table 5. Principal component analysis results show that two independent principal components can be selected as evaluation indicators of student academic performance, explaining 82.06% of the total variance (cumulative contribution rate), essentially representing the vast majority of information regarding students' grades in the eight courses. Furthermore, principal component analysis outputs the factor scores (i.e., the comprehensive indicators Fi) of each student's two principal components (i.e., the scores of each comprehensive indicator Fi). kj (k = 1 to 55, j = 1 or 2), the results are shown in Table 6.

[0050] Table 5

[0051]

[0052] (2) Comprehensive Evaluation D Value: Calculate the membership function value and weight of each comprehensive indicator for each student to obtain the comprehensive evaluation value (D) of each student's academic performance. The membership function method is used to calculate the membership function values ​​of the academic performance of 8 courses for 55 students and calculate their weighted average to evaluate the comprehensive academic performance of the 55 students. The membership function value of each principal component factor score (i.e., each comprehensive indicator) for each student is calculated using formula (4), where L kj F is the j-th comprehensive index of student k. kj The membership function value; F kj F represents the j-th comprehensive index value of student k; max(j) and F min(j) These are the maximum and minimum values ​​of the j-th comprehensive index for all students. The results are shown in Table 6.

[0053]

[0054] The weights of each membership function value are calculated using formula (5) based on the contribution rate of each principal component, where W j P represents the importance of the j-th composite indicator among all composite indicators; j This represents the contribution rate of the j-th comprehensive indicator, where j = 1 or 2. The results are shown in Table 6.

[0055]

[0056] The comprehensive evaluation value is calculated using formula (6), where D k The value is the comprehensive evaluation value of student k's academic performance, where k = student ID; j = 1 or 2.

[0057] D k =∑[L kj ·W j (6).

[0058] (3) Classification: The K-means clustering method was used to cluster the D values ​​of each student. The number of clusters was set to 5, that is, the D values ​​were divided into 5 categories, which correspond to the overall academic performance of A (excellent), B (good), C (medium), D (poor) and E (very poor) from large to small. The results are shown in Table 6.

[0059] Table 6

[0060]

[0061] Comparison of Example 1 and Comparative Example 1

[0062] Example 1 and Comparative Example 1 used different methods to comprehensively evaluate the academic performance of 55 students in 8 courses. The cov() function in R language was used to calculate the evaluation value Z for each student according to the method in Example 1.k (Table 4) The evaluation score D for each student calculated according to the method of Comparative Example 1. k (Table 6) Correlation analysis showed a correlation coefficient r = 0.9959, indicating a highly significant positive correlation.

[0063] The number of people in each of the five levels in the comprehensive evaluation results using the method of Example 1 and the method of Comparative Example 1 were counted separately, and the percentage of people in each level was calculated, as shown in Table 7.

[0064] Table 7

[0065]

[0066] For the number of students appearing in each level of the two methods in Table 7, the wilcox.test() function in R language was called. After a nonparametric hypothesis test, the result of the wilcoxon rank test was w=0, p-value=99.99>0.05, indicating that there is no significant difference in the results of evaluating students' academic performance levels using the method of Example 1 and the method of Comparative Example 1.

[0067] The results above indicate that the number of people to be evaluated in each level is relatively consistent, and the overall trend is also the same.

[0068] Advantages of the method in Example 1: (1) Each step of the method in Example 1 has a clear meaning, distinct logic, and simple calculation. Its results are consistent with the membership function method. Moreover, the method in Example 1 can be implemented using Excel, SPSS, or R language, making it simpler, more intuitive, and easier to understand than existing methods. (2) Indicators for evaluating student learning performance, such as scores in mathematics, Chinese, and English, generally show that the larger the indicator value, the better the learning performance, exhibiting a positive trend. If there is no positive trend, a reciprocal transformation is required. For example, some indicators show that the smaller the better, so they are generally transformed inversely to become the larger the better, so that they have a positive trend with other indicators that are better the larger. (3)

[0069] In Example 1, the minimum value of the transformed data for each variable is 0, which represents the worst-case scenario for a student's performance in a particular course. Therefore, the student's overall academic performance (Z) can be determined by calculating the degree to which the student's performance in all courses deviates from their worst-case scenario. k The greater the deviation, the better the overall performance; that is, the closer to 1, the better the overall performance of the student being evaluated.

[0070] Example 2

[0071] In October 2022, 81 wheat varieties were planted in the dry and infertile experimental field of the Xihuayuan Experimental Station of Cangzhou Academy of Agricultural and Forestry Sciences in the Heilonggang River Basin. The randomized block design was used with 3 replicates. The wheat was irrigated once during the greening period (one irrigation in spring). Other field management was the same as that of ordinary fields. In June 2023, the plant height, ear and grain characteristics of each wheat variety were investigated. See Table 8 for details.

[0072] Table 8

[0073]

[0074]

[0075] (1) Correlation analysis: Correlation coefficients were used to determine the pairwise correlations between 10 trait indicators of 81 wheat varieties. Let X be the correlation coefficient. ik or X i'k The variables or observed values ​​representing traits are: i or i', where i or i' is the trait index code, i or i' = 1 represents plant height, i or i' = 2 represents single ear weight, i or i' = 3 represents ear length, i or i' = 4 represents the number of spikelets per ear, i or i' = 5 represents the number of grains per ear, i or i' = 6 represents the thousand-grain weight, i or i' = 7 represents the grain area, i or i' = 8 represents the grain perimeter, i or i' = 9 represents the grain length, and i or i' = 10 represents the grain width; k is the wheat variety code, which is a number from 1 to 81. Taking the correlation between plant height and single ear weight of 81 wheat varieties as an example, the correlation coefficient between plant height and single ear weight is calculated based on formula (1), where r is the correlation coefficient. Let i be the average of the trait indices of all wheat varieties. The trait index of all wheat varieties is the average of i'. When i = 1 and k = 1, the plant height of wheat with variety code 1 is X. 11 When i' = 2 and k = 1, the weight of a single ear of wheat with variety code 1 is X. 21 , The average plant height of 81 wheat varieties. This represents the average single ear weight of 81 wheat varieties, with k ranging from 1 to 81.

[0076]

[0077] The results of a simple correlation analysis of 10 trait indicators of 81 wheat varieties are shown in Table 9.

[0078] Table 9

[0079]

[0080] Note: * indicates a significant difference at the 0.05 level; ** indicates an extremely significant difference at the 0.01 level.

[0081] The results in Table 9 show that, under the condition of one spring irrigation in dry and thin land in the Heilonggang watershed, 10 indicators of plant height, ear size and grain characteristics of 81 wheat varieties showed different degrees of correlation, and all of them showed the same positive trend as the overall performance of wheat. Therefore, no data transformation was required, and the data in this table can be used as the basis for comprehensive evaluation.

[0082] (2) Calculate the converted values ​​based on formula (2).

[0083]

[0084] Among them, u ik For X ik The converted value; X min(i) For 81 X i The minimum value of X max(i) For 81 X i The maximum value in; i ranges from 1 to 10, and k ranges from 1 to 81.

[0085] The converted numerical results are shown in Table 10.

[0086] (3) Calculate the comprehensive evaluation value based on formula (3).

[0087]

[0088] Among them, Z k The comprehensive evaluation value for each wheat variety, u ik For X ik The converted value, u min(i) For 81 u i The minimum value in the range is n, where n is the total number of all i, which is 10 in this case. The results are shown in Table 11.

[0089] (4) Classification by level: Z k Cluster analysis was performed on the evaluation values ​​Z to achieve grade classification. The K-means clustering method was used to evaluate the Z values ​​of 81 wheat varieties. k Perform clustering, setting the number of clusters to 5, i.e., Z0 k The values ​​are divided into 5 categories, corresponding to a comprehensive performance level of A (Excellent), B (Good), C (Average), D (Poor), and E (Very Poor) from highest to lowest, and are represented by a comprehensive evaluation value Z. k The wheat varieties were ranked according to their overall performance evaluation values, as shown in Table 11.

[0090] Table 10

[0091]

[0092]

[0093] Table 11

[0094]

[0095]

[0096] Table 11 shows that varieties such as Cangmai 6002, Cangmai 6005, Cangmai 6003, Jimai 32, and Nayan 101 exhibit strong drought and salt tolerance. Under the arid and infertile conditions of the Heilonggang watershed with only one spring irrigation, it is suitable to plant wheat varieties with strong stress resistance, hence their relatively high ranking in the comprehensive evaluation. Gaoyou 8901, Gaoyou 5218, Gaoyou 5766, and Shilu 02-1 are high-quality wheat varieties, but their field performance and yield under the arid and infertile conditions of the Heilonggang watershed with only one spring irrigation are average, resulting in a relatively low ranking in the comprehensive evaluation. Therefore, this evaluation result is largely consistent with the field performance and practical experience of wheat.

[0097] Comparative Example 2

[0098] The membership function method is used for comprehensive evaluation. The specific calculation method and process are as follows.

[0099] (1) Principal Component Analysis: Principal component analysis was performed on 10 indicators of plant height, spike size, and grain characteristics of 81 wheat varieties. The principal component factors obtained from the principal component analysis were used as comprehensive indicators to evaluate the field performance of each variety. Three independent principal components were selected as the comprehensive indicators of the field performance of the varieties for overall judgment. The princomp() function in R language was used to perform principal component analysis on the 10 trait indicators of the 81 varieties to obtain the eigenvalues ​​and contribution rates P of each principal component. j (j = 1, 2, or 3), see Table 12. Principal component analysis results show that three independent principal components can be selected as evaluation indicators for variety field performance, explaining 83.07% of the total variation (cumulative contribution rate), essentially representing most of the information from the 10 original indicators. Furthermore, principal component analysis outputs the factor scores (i.e., the comprehensive indicators Fi) of the three principal components (1, 2, and 3) for each variety. kj (k = 1 to 81, j = 1, 2, or 3).

[0100] The results are shown in Table 13.

[0101] Table 12

[0102]

[0103] (2) Comprehensive evaluation D value: Calculate the membership function value and weight of each comprehensive index for each wheat variety, and then obtain the comprehensive evaluation value (D) of the field performance of each wheat variety. The membership function method is used to calculate the membership function values ​​of 10 indicators for 81 varieties and calculate their weighted average to evaluate the comprehensive field performance of 81 varieties. The membership function value of each principal component factor score (i.e., each comprehensive index) of each wheat variety is calculated by formula (4).

[0104] The results are shown in Table 13.

[0105]

[0106] Among them, L kj F is the j-th comprehensive index of wheat variety k. kj The membership function value; F kj F is the j-th comprehensive index value of variety k; max(j) and F min(j) These are the maximum and minimum values ​​of the j-th comprehensive index for all varieties, respectively.

[0107] The weight of each membership function value is calculated using formula (5) based on the contribution rate of each principal component.

[0108] The results are shown in Table 13.

[0109]

[0110] Among them, W j P represents the importance of the j-th composite indicator among all composite indicators; j This represents the contribution rate of the j-th comprehensive indicator, where j = 1, 2, or 3.

[0111] The comprehensive evaluation value is calculated using formula (6).

[0112] D k =∑[L kj ·W j (6).

[0113] Among them, D k The value is the comprehensive evaluation value of the field performance of wheat variety k, where k = variety code; j = 1, 2, or 3.

[0114] (3) Grading: The K-means clustering method was used to cluster the D values ​​of each variety. The number of clusters was set to 5, that is, the D values ​​were divided into 5 categories, which correspond to the comprehensive performance of A (excellent), B (good), C (medium), D (poor) and E (very poor) from large to small.

[0115] Table 13

[0116]

[0117]

[0118] Comparison of Example 2 and Comparative Example 2

[0119] Example 2 and Comparative Example 2 used different methods to comprehensively evaluate the field performance of 81 wheat varieties. The cov() function in R language was used to calculate the evaluation value Z for each wheat variety according to the method in Example 2. k (Table 11) The evaluation value D for each wheat variety calculated according to the method of Comparative Example 2. k (Table 13) Correlation analysis showed a correlation coefficient r = 0.9837, indicating a highly significant positive correlation.

[0120] The number of varieties in the five levels of the comprehensive evaluation results using the method of Example 2 and the method of Comparative Example 2 were counted separately, and the percentage of wheat varieties in each level was calculated, as shown in Table 14.

[0121] Table 14

[0122]

[0123] For the materials appearing at each level in the two methods in Table 14, the wilcox.test() function in R language was called. After nonparametric hypothesis testing, the result of the wilcoxon rank test was w = -0.135, p-value = 0.8927 > 0.05, indicating that there is no significant difference between the results of the evaluation of the overall field performance of wheat varieties by the method of Example 2 and the method of Comparative Example 2.

[0124] The results above indicate that the apparent number of materials to be evaluated in each level is relatively consistent, and the overall trend of change is also the same.

[0125] Advantages of the method in Example 2: (1) Each step of the method in Example 2 has a clear meaning, distinct logic, and simple calculation. Its results are consistent with the membership function method. Moreover, the method in Example 2 can perform related calculations through Excel, SPSS, or R language, which is simpler, more intuitive, and easier to understand than existing technical methods. (2) The indicators for evaluating the overall field performance of wheat varieties, such as plant height, spike length, and thousand-grain weight, all show that the larger the indicator value, the better the field performance, exhibiting a positive trend. If there is no positive trend, corresponding transformation is required. For example, some indicators show that the smaller the better, so they are generally transformed inversely to become the larger the better, so that they have a positive trend with other indicators that have the larger the better. (3) The minimum value of the data after the transformation of each variable in the method in Example 2 is 0, which is the worst state of the field performance of a certain indicator of wheat variety. Therefore, the overall field performance (Z) of the material to be evaluated can be determined by solving the degree to which all indicators deviate from their worst state. k The greater the deviation, the better the overall performance; that is, the closer to 1, the better the overall performance of the material being evaluated.

Claims

1. A multivariate evaluation method, comprising the following steps: 1) Perform correlation analysis on multiple indicators of the object to be evaluated, remove the indicators of the object to be evaluated that have no significant correlation with other indicators, and retain the indicators of the object to be evaluated that have significant correlation, thus obtaining the retained indicators of the object to be evaluated. 2) If the changing trends of all the indicators in the retained indicators are positive, then there is no need to take the reciprocal value of any of the retained indicators; if there are any indicators in the retained indicators with a negative changing trend, then the reciprocal value of the indicator is taken. After this step of positive homotrification, the positive homotrification values ​​of the indicators of the object to be evaluated are obtained. 3) Convert the positive trend values ​​of the indicators of the object to be evaluated into indicator conversion values ​​in the range of [0,1]. 4) Determine the number n of the positive trend indicators of the object to be evaluated, and then calculate the average of the sum of the deviations of the n indicator conversion values ​​of the object to be evaluated from the minimum value of the corresponding indicator conversion values ​​of all objects to be evaluated, so as to obtain the evaluation value used to evaluate the object to be evaluated. 5) Compare the evaluation value of the object to be evaluated with 1. The closer the evaluation value is to 1, the better the overall performance of the object to be evaluated.

2. The multivariate evaluation method according to claim 1, characterized in that, In step 1), the correlation coefficient is used to determine the pairwise correlation between multiple indicators of the object to be evaluated, and the correlation coefficient is calculated based on formula 1): Where r is the correlation coefficient, i or i' is the indicator, k is the object to be evaluated, and X ik Let i be the positive trend-normalized value of index i for the object k to be evaluated. X is the average of the positively trending values ​​of indicator i for all objects to be evaluated. i'k Let i' be the positive trend-based value of the index i' of the object to be evaluated k. The average of the positively trending values ​​of indicator i' of all objects to be evaluated.

3. The multivariate evaluation method according to claim 1, characterized in that, In step 3), the converted value of the index is calculated based on formula 2): Among them, u ik For X ik The converted value of the indicator; X min(i) For X among all objects to be evaluated i The minimum value of X max(i) For X among all objects to be evaluated i The maximum value.

4. The multivariate evaluation method according to claim 1, characterized in that, In step 4), the evaluation value is calculated based on formula 3): Among them, Z k u is the comprehensive evaluation value of the object to be evaluated. ik For X ik The indicator conversion value, u min(i) For the object to be evaluated, u i The minimum value of , where n is the number of positive homotrend indicators of the object to be evaluated.

5. The multivariate evaluation method according to claim 1, characterized in that, It also includes step 6), based on Z in step 4). k Perform clustering or sorting.

6. The multivariate evaluation method according to claim 5, characterized in that, When there are N objects to be evaluated, the K-means clustering method is used to cluster the Z-axis of the objects to be evaluated. k Perform clustering and set the number of clusters L, where L < N.

Images