Personnel performance evaluation method based on IWOA-SVM

Abstract

The invention belongs to the technical field of machine learning models, particularly relates to a personnel performance evaluation method based on IWOA-SVM, and solves the problems that a traditional support vector machine (SVM) is low in precision, difficult in parameter selection and the like in performance intelligent evaluation. The method comprises the steps that Tent chaotic mapping and a pseudo-opposition learning strategy are utilized to increase the diversity and quality of an initial population, and the whale algorithm (WOA) is prevented from falling into local optimum; the global optimization capability of the WOA is improved by adopting a differential evolution mechanism; a penalty factor and kernel function parameters of the SVM are optimized through an improved whale algorithm (IWOA), and performance evaluation can be effectively carried out while optimal parameters are obtained. According to the method, the whale algorithm can be improved by using Tent chaotic mapping, pseudo-opposition learning and a differential evolution strategy, SVM parameters are searched in a global range, and better model performance is obtained.

Description

Technical Field

[0001] This invention relates to the field of machine learning models, and specifically to a personnel performance evaluation method based on IWOA-SVM. Background Technology

[0002] Special equipment, as crucial components related to national economic development and public safety, is applied in all aspects of production and daily life. With the development of the social economy, its quantity and parameters are increasing, posing higher demands on the safety supervision of special equipment. In the current special equipment management model, the safe operation of special equipment mainly relies on the professional competence of operators and inspection personnel. While special equipment inspection agencies have established corresponding management systems and regulatory departments to manage the work efficiency of their inspection personnel, issues such as differences in the work abilities of inspection personnel, mismatches between professional competence and required skills, and a lack of enthusiasm in their work pose significant challenges to the management of inspection agencies. Therefore, an effective method is needed to assess and evaluate the work ability, performance, and attitude of inspection personnel.

[0003] Performance, as a metric, encompasses an individual's describable work behaviors and measurable work results within a specific timeframe. It also includes the organization's assessment of the individual's past qualities and abilities, guiding improvement and ultimately predicting the individual's future performance within a specific timeframe. Applying performance appraisal to the management of inspection agencies can improve the work capabilities and attitudes of inspection personnel and reduce the error rate in special equipment inspections. Traditional performance appraisals for inspection and testing personnel rely primarily on manual processes by relevant department staff, using reports to record personnel's performance. This method suffers from limitations such as a lack of scientific rigor and subjective results. Therefore, a standardized performance appraisal system for inspection and testing personnel needs to be constructed using modern mathematical algorithms.

[0004] In recent years, with the rapid development of artificial intelligence technology, machine learning methods have been widely applied to performance appraisal. Researchers input performance evaluation indicators into machine learning models to achieve intelligent judgment in performance evaluation. Commonly used machine learning models include Support Vector Machines (SVM), Random Forests, Naive Bayes, and Artificial Neural Networks. Among them, SVM has a sound theoretical foundation and unique advantages in handling small-sample learning problems. However, the random selection of penalty factors and kernel function parameters in SVM can easily affect classification and recognition performance. Summary of the Invention

[0005] The purpose of this invention is to solve the problems in the background art and provide a personnel performance evaluation method based on IWOA-SVM.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0007] A personnel performance evaluation method based on IWOA-SVM includes:

[0008] S1. Data Preprocessing: Through Min-Max normalization, the original data of performance indicators (such as workload, error rate, collaboration ability, etc.) are mapped to 0,1 comparable data, thereby dynamically quantifying the indicators, solving the problem of integrating indicators of different dimensions, and providing standardized input for support vector machine (SVM).

[0009] S2. Divide the entire dataset into training and test sets: the training set is used to train the SVM model, and the test set is used to evaluate the performance of the SVM model.

[0010] S3. Constructing the IWOA-SVM model: ① Introducing Tent chaotic mapping, pseudo-oppositional learning, and differential evolution mechanisms into the traditional Whale Algorithm (WOA) to construct an improved Whale Algorithm (IWOA); ② Setting control parameters in the improved Whale Algorithm (IWOA): convergence factor a, behavior selection coefficient A, random perturbation coefficient C, spiral shape parameter l, and behavior probability selection parameter p; ③ Optimizing the penalty factor c and kernel function parameter g of the SVM using the improved Whale Algorithm (IWOA) to form the IWOA-SVM model, i.e., using the parameters (c, g) as the position iteration calculation for each whale individual in the improved Whale Algorithm (IWOA); After data initialization, N whale individuals are randomly generated as the initial whale group, with a maximum number of iterations. The location of each individual whale For each individual whale, calculate its fitness value, which is the value of the objective function; select the n whales with the smallest fitness values ​​as the initial population positions, and record the optimal fitness value and its corresponding position; then, update the control parameters; ④ use the training set to train, thereby obtaining the optimal parameters (c, g);

[0011] S4. Obtaining intelligent evaluation results based on optimal SVM parameters: Using the optimized optimal parameters (c, g), construct the final SVM classification model, and use the test set to evaluate its performance classification performance to obtain the final performance evaluation results.

[0012] Preferably, in step S1, the Min-Max normalization process is calculated using the following formula: Normalized data express, It is normalized data. These are the maximum and minimum values ​​in the current data, respectively.

[0013] Preferably, in step S3, the penalty factor c comes from the original form of SVM, and the kernel function parameter g comes from the SVM decision function;

[0014] The original form of the SVM formula is: ,in, Represents slack variables. Indicates that the first The interval of deviation of each data point As a penalty factor;

[0015] The IWOA-SVM model uses an SVM decision function that incorporates a Lagrange function based on the original SVM form. and Gaussian kernel function ,in, For Lagrange factors, For Gaussian kernel function, These are the parameters for the kernel function.

[0016] Preferably, in step S3, the Tent mapping method is as follows:

[0017]

[0018] After generating chaotic values, the mapping rule between chaotic values ​​and the population space is as follows:

[0019]

[0020] In the formula: Let i be the position of individual i in dimension j; and Let i be the position boundary of individual i in dimension j; d is the position dimension.

[0021] Preferably, in step S3, the pseudo-oppositional learning method is as follows: Let the individual whale... Location X i = ( x i , 1 , x i , 2 ,..., x i , d ), x i , j ∈ [ lb j , ub j ] Individual whale The pseudo-opposing whale individual is Defined as:

[0022]

[0023]

[0024] In the formula: , For position The opposite position, This means taking a random value between p and q.

[0025] Preferably, in step S3, during the iterative calculation of the whale's individual position, the convergence factor 'a' is a coefficient that decays from 2 to 0 as the iteration begins, depending on the current iteration number. Calculate, let the current iteration number be... , To determine the maximum number of iterations, we have: The behavior selection coefficient A and the random perturbation coefficient C are coefficients used to control whale behavior, and are set as follows: For a random vector in the range [0,1], then: ; .

[0026] Preferably, in step S3, during the iterative calculation of the position of the individual whale, the probability selection parameter p is a random number uniformly distributed in the range [0,1].

[0027] when At that time, and At this time, an encirclement strategy is adopted:

[0028]

[0029]

[0030] The position of the current best solution is represented as The location of an individual whale is represented as The distance between an individual whale and its prey is , Step size;

[0031] when At that time, and At that time, a spiral update strategy was adopted, with individual whales and... prey Distance between:

[0032]

[0033]

[0034] in, The distance from an individual whale to its prey. A constant is used to define the shape of the logarithmic spiral, and the spiral shape parameter is... A random number in the range [-1, 1].

[0035] In conclusion, when The overall mathematical model is described as follows:

[0036] (1)

[0037] Preferably, in step S3, during the iterative calculation of the position of the individual whale, when At that time, the whale updates its position based on randomly selected individuals, rather than the position of the optimal individual. The update formula is as follows:

[0038] (2)

[0039] (3)

[0040] The position vector of a randomly selected individual whale from the group is represented as follows: The distance from a randomly selected individual whale to its prey is represented as... .

[0041] Preferably, in step S3, during the iterative calculation of the whale's individual position, the differential evolution mechanism utilizes a mutation expansion factor. and crossover probability Effectively enhances the global search capability of intelligent algorithms, specifically including mutation, crossover, and greedy selection;

[0042] ①The mutation operator is defined as follows:

[0043] (4)

[0044] In the formula: v is the mutated solution, x is the current solution, r1, r2, and r3 represent random integers in [1, N], corresponding to three independent individuals, and SF is the mutation expansion factor, which is used to control the degree of change between the original individual and the mutated individual;

[0045] ② Crossover operator: After individual mutation, the crossover operator can be used to further enhance population diversity on the target vector. Differential evolution uses the crossover operator on xi and vi to generate offspring individuals zi, specifically defined as:

[0046] (5)

[0047] In the formula: rand represents a random value between [0,1], and CR represents the crossover probability of individuals;

[0048] ③ Selection operator: This operator uses a greedy strategy to compare the parent individual with its child individual to select the new individual. Specifically, it is defined as follows:

[0049] (6)

[0050] In the formula: f(z) is the fitness function value of z, defined as follows:

[0051]

[0052] Accuracy is the proportion of correctly classified samples out of the total number of samples, as shown below:

[0053]

[0054] In the formula: TP represents true positives, FP represents false positives, TN represents true negatives, and FN represents false negatives. The selection probability Pi determines whether an individual whale updates its position according to the differential evolution strategy. The selection probability of xi is defined as the ratio between the fitness of xi and the overall fitness of the population, expressed as:

[0055]

[0056] Based on the selection probability Pi, the individual position is updated using the calculation method for the current solution xi. The specific method is as follows:

[0057] like If rand is a random number in [0,1], then the current solution xi is updated in position using equations (1)-(3); if Then the current solution xi is updated in position using equations (4)-(6).

[0058] In summary, the beneficial effects of this invention are as follows:

[0059] 1. The personnel performance evaluation method based on IWOA-SVM described in this invention improves the objectivity of the evaluation: Tent chaotic mapping and pseudo-opposition learning ensure that the initial population covers diverse performance scenarios and reduces subjective bias; the optimal parameters of SVM are automatically generated, eliminating the arbitrariness of manual parameter tuning.

[0060] 2. The personnel performance evaluation method based on IWOA-SVM described in this invention has enhanced dynamic adaptability: the differential evolution mechanism responds to changes in indicators in real time and adapts to the iterative requirements of the job; the probability selection strategy dynamically adjusts the optimization path to match different evaluation cycles.

[0061] 3. The personnel performance evaluation method based on IWOA-SVM described in this invention uses Tent chaotic mapping and pseudo-oppositional learning strategies to increase the diversity and quality of the initial population, avoiding the Whale Algorithm (WOA) from getting trapped in local optima; it adopts a differential evolution mechanism to improve the global optimization ability of WOA; and it optimizes the penalty factor and kernel function parameters of SVM through the improved Whale Algorithm (IWOA), so as to effectively evaluate performance while obtaining the optimal parameters. Attached Figure Description

[0062] Figure 1 This is a flowchart of the IWOA-SVM algorithm in this invention. Detailed Implementation

[0063] The following specific embodiments are merely illustrative of the present invention and are not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to these embodiments without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of the present invention.

[0064] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0065] Example 1:

[0066] according to Figure 1 As shown, a personnel performance evaluation method based on IWOA-SVM includes:

[0067] S1. Data Preprocessing: Through Min-Max normalization, the original performance indicator data (such as workload, error rate, collaboration ability, etc.) are mapped to 0,1 comparable data, thereby dynamically quantifying the indicators, solving the problem of integrating indicators with different dimensions, and providing standardized input for Support Vector Machine (SVM); in step S1, the Min-Max normalization processing is calculated using the following formula: Normalized data express, It is normalized data. These are the maximum and minimum values ​​in the current data, respectively.

[0068] S2. Divide the entire dataset into training and test sets: the training set is used to train the SVM model, and the test set is used to evaluate the performance of the SVM model.

[0069] S3. Construct the IWOA-SVM model:

[0070] ① An improved whale algorithm (IWOA) is constructed by introducing Tent chaotic mapping, pseudo-oppositional learning and differential evolution mechanism into the traditional whale algorithm (WOA).

[0071] The Tent mapping method is as follows:

[0072]

[0073] After generating chaotic values, the mapping rule between chaotic values ​​and the population space is as follows:

[0074]

[0075] In the formula: Let i be the position of individual i in dimension j; and Let i be the position boundary of individual i in dimension j; d is the position dimension.

[0076] The pseudo-oppositional learning method is as follows: Let the individual whale... Location X i = ( x i , 1 , x i , 2 ,..., x i , d ), x i , j ∈ [ lb j , ub j ] Individual whale The pseudo-opposing whale individual is Defined as:

[0077]

[0078]

[0079] In the formula: , For position The opposite position, This means taking a random value between p and q.

[0080] ②The improved whale algorithm (IWOA) sets the following control parameters: convergence factor a, behavior selection coefficient A, random perturbation coefficient C, spiral shape parameter l, and behavior probability selection parameter p;

[0081] ③ The penalty factor c and kernel function parameter g of the SVM are optimized using the Improved Whale Algorithm (IWOA) to form the IWOA-SVM model. Specifically, the parameters (c, g) are used as the position of each individual whale in the IWOA algorithm for iterative calculation. After data initialization, N individual whales are randomly generated as the initial whale group, with a maximum number of iterations. The location of each individual whale For each individual whale, its fitness value, i.e. the value of the objective function, is calculated; the n whales with the smallest fitness values ​​are selected as the initial population positions, and the optimal fitness value and its corresponding position are recorded; then, the control parameters are updated.

[0082] The penalty factor c comes from the original form of SVM, and the kernel function parameter g comes from the SVM decision function;

[0083] The original form of the SVM formula is: ,in, Represents slack variables. Indicates that the first The interval of deviation of each data point As a penalty factor;

[0084] The IWOA-SVM model uses an SVM decision function that incorporates a Lagrange function based on the original SVM form. and Gaussian kernel function ,in, For Lagrange factors, For Gaussian kernel function, These are the parameters for the kernel function.

[0085] During the iterative calculation of the whale's position, the convergence factor 'a' is a coefficient that decays from 2 to 0 as the iteration begins, depending on the current iteration number. Calculate, let the current iteration number be... , To determine the maximum number of iterations, we have: The behavior selection coefficient A and the random perturbation coefficient C are coefficients used to control whale behavior, and are set as follows: For a random vector in the range [0,1], then: ; .

[0086] The probability selection parameter p is a random number uniformly distributed in the range [0,1].

[0087] when At that time, and At this time, an encirclement strategy is adopted:

[0088]

[0089]

[0090] The position of the current best solution is represented as The location of an individual whale is represented as The distance between an individual whale and its prey is , Step size;

[0091] when At that time, and At that time, a spiral update strategy was adopted, with individual whales and... prey Distance between:

[0092]

[0093]

[0094] in, The distance from an individual whale to its prey. A constant is used to define the shape of the logarithmic spiral, and the spiral shape parameter is... A random number in the range [-1, 1].

[0095] In conclusion, when The overall mathematical model is described as follows:

[0096] (1)

[0097] when At that time, the whale updates its position based on randomly selected individuals, rather than the position of the optimal individual. The update formula is as follows:

[0098] (2)

[0099] (3)

[0100] The position vector of a randomly selected individual whale from the group is represented as follows: The distance from a randomly selected individual whale to its prey is represented as... .

[0101] Furthermore, differential evolution mechanisms utilize mutation expansion factors. and crossover probability Effectively enhances the global search capability of intelligent algorithms, specifically including mutation, crossover, and greedy selection;

[0102] ①The mutation operator is defined as follows:

[0103] (4)

[0104] In the formula: v is the mutated solution, x is the current solution, r1, r2, and r3 represent random integers in [1, N], corresponding to three independent individuals, and SF is the mutation expansion factor, which is used to control the degree of change between the original individual and the mutated individual;

[0105] ② Crossover operator: After individual mutation, the crossover operator can be used to further enhance population diversity on the target vector. Differential evolution uses the crossover operator on xi and vi to generate offspring individuals zi, specifically defined as:

[0106] (5)

[0107] In the formula: rand represents a random value between [0,1], and CR represents the crossover probability of individuals;

[0108] ③ Selection operator: This operator uses a greedy strategy to compare the parent individual with its child individual to select the new individual. Specifically, it is defined as follows:

[0109] (6)

[0110] In the formula: f(z) is the fitness function value of z, defined as follows:

[0111]

[0112] Accuracy is the proportion of correctly classified samples out of the total number of samples, as shown below:

[0113]

[0114] In the formula: TP represents true positives, FP represents false positives, TN represents true negatives, and FN represents false negatives. The selection probability Pi determines whether an individual whale updates its position according to the differential evolution strategy. The selection probability of xi is defined as the ratio between the fitness of xi and the overall fitness of the population, expressed as:

[0115]

[0116] Based on the selection probability Pi, the individual position is updated using the calculation method for the current solution xi. The specific method is as follows:

[0117] like If rand is a random number in [0,1], then the current solution xi is updated in position using equations (1)-(3); if Then the current solution xi is updated in position using equations (4)-(6).

[0118] ④ Use the training set to train and obtain the optimal parameters (c, g);

[0119] S4. Obtaining intelligent evaluation results based on optimal SVM parameters: Using the optimized optimal parameters (c, g), construct the final SVM classification model, and use the test set to evaluate its performance classification performance to obtain the final performance evaluation results.

[0120] Example 2:

[0121] Examples of performance appraisals for special equipment testing and inspection personnel

[0122] I. Data Collection and Indicator System Construction

[0123] 1. Evaluation Subjects

[0124] Fifty front-line inspection personnel from a special equipment inspection and testing organization.

[0125] 2. Performance Indicator System

[0126] A quantitative evaluation system comprising 4 primary indicators and 10 secondary indicators is established, as shown in the table below:

[0127]

[0128] 3. Performance rating labeling

[0129] Based on their overall performance, the expert committee divided the 50 inspectors into three performance levels: Excellent (Level 3): 15 people, Satisfactory (Level 2): ​​25 people, and Needs Improvement (Level 1): 10 people.

[0130] II. Data Preprocessing

[0131] 1. Data normalization

[0132] Min-Max normalization is used to map each indicator data to the [0,1] interval:

[0133] Raw data: highest 98%, lowest 75%, one person's score was 92%.

[0134] x_new=(92-75) / (98-75)=17 / 23≈0.739

[0135] 2. Dataset partitioning

[0136] Training group: 35 people (10 excellent + 18 qualified + 7 needing improvement)

[0137] Test set: 15 people (5 excellent, 7 satisfactory, and 3 needing improvement)

[0138] III. IWOA-SVM Model Parameter Settings

[0139] 1. SVM parameter search range

[0140] Penalty factor c: [0.1, 100]

[0141] Kernel function parameter g: [0.001, 10]

[0142] 2. IWOA Algorithm Parameters

[0143] Population size N=30

[0144] Maximum number of iterations T_max=100

[0145] Differential evolution parameters: SF=0.5, CR=0.9

[0146] Helical shape constant b=1

[0147] IV. Example Demonstration of IWOA Optimization Process

[0148] 1. First iteration:

[0149] The convergence factor a = 1.98 is calculated.

[0150] Randomly generated: r=0.65, l=-0.3, p=0.42

[0151] Calculated coefficients: A = 0.594, C = 1.3

[0152] Since A < 1 and p = 0.42 < 0.5, an encirclement strategy is adopted.

[0153] The calculated SVM parameters are: X = (c = 15.3, g = 0.85). Using these parameters to train the SVM, the accuracy on the training set is 82.9%, and the fitness is 0.171.

[0154] 2. 50th iteration

[0155] Calculate the convergence factor a = 1.0

[0156] Randomly generated: r=0.25, p=0.67

[0157] A=0.5<1, p=0.67>0.5, a spiral update strategy is adopted.

[0158] The calculated SVM parameters are: X=(c=8.7, g=1.2), accuracy=91.4%, and the individual updates its position according to the spiral formula, moving closer to the optimal solution.

[0159] 3. 100th iteration (final)

[0160] The global optimal solution was found. SVM parameters: X=c=12.5, g=0.95.

[0161] Accuracy on the training set: 93.6%

[0162] V. Model Performance Evaluation

[0163] Validate on the test set using the optimal parameters:

[0164] 1. Confusion matrix results

[0165] Forecast / Actual Needs improvement (1) Qualified(2) Excellent (3) Needs improvement (1) 2 1 0 Qualified(2) 0 6 4 Excellent (3) 1 0 4

[0166] 2. Performance index calculation:

[0167] Overall accuracy: (2+6+4) / 15=80.0%

[0168] Accuracy:

[0169] Needs improvement: 2 / (2+1+0) = 66.7%

[0170] Passed: 6 / (0+6+1)=85.7%

[0171] Excellent: 4 / (1 + 0 + 4) = 80.0%

[0172] Recall rate:

[0173] Needs improvement: 2 / 3 = 66.7%

[0174] Passed: 6 / 7=85.7%

[0175] Excellent: 4 / 5 = 80.0%

[0176] VI. Comparison with traditional methods

[0177] Evaluation methods accuracy objectivity efficiency Explainability Human subjective evaluation 65-75% Difference Low middle Traditional SVM 85.3% good middle good IWOA-SVM 93.6% excellent high good

[0178] VIII. Summary of Implementation Examples

[0179] This embodiment fully demonstrates the practical application process of the IWOA-SVM method in the performance evaluation of special equipment inspection personnel. By automatically finding the optimal SVM parameters through an intelligent optimization algorithm, this method achieves: significantly improved evaluation accuracy (8-28% higher than traditional methods), completely objective evaluation, elimination of human bias, efficient automated processing, and a substantial reduction in management costs, providing a scientific basis for precise human resource management.

Claims

1. A personnel performance evaluation method based on IWOA-SVM, characterized in that, Comprise: S1, data preprocessing: through Min-Max normalization, the performance index original data is mapped to 0, 1 comparable data, so as to dynamically quantify the index, solve the fusion problem of different dimension index, provide standardized input for support vector machine (SVM); S2, the whole data set is divided by training set and test set: the training set is used to train SVM model, and the test set is used to evaluate the performance of SVM model; S3, constructing an IWOA-SVM model: ① introducing Tent chaotic mapping, pseudo opposition learning and differential evolution mechanism into a traditional whale optimization algorithm (WOA) to construct an improved whale optimization algorithm (IWOA); ② setting control parameters in the improved whale optimization algorithm (IWOA), including a convergence factor a, a behavior selection coefficient A, a random disturbance coefficient C, a spiral shape parameter l and a behavior probability selection parameter p; ③ using the improved whale optimization algorithm (IWOA) to optimize a penalty factor c and a kernel function parameter g of SVM to form an IWOA-SVM model, that is, taking the parameters (c, g) as the position of each whale individual in the improved whale optimization algorithm (IWOA) for iterative calculation; after data initialization, generating N whale individuals in a random manner as an initial whale population, and setting a maximum iteration number , the position of each whale individual , for each whale individual, calculating a fitness value, that is, a value of a target function; selecting n whales with the minimum fitness value as an initial population position, and recording the optimal fitness value and the corresponding position; then, updating the control parameters; ④ training using a training set to obtain optimal parameters (c, g); S4, the intelligent evaluation result based on the optimal SVM parameter is obtained: using the optimized optimal parameter (c, g), the final SVM classification model is constructed, and its performance classification performance is evaluated using the test set, and the final performance evaluation result is obtained.

2. The personnel performance evaluation method based on IWOA-SVM according to claim 1, wherein, In the step S1, the Min-Max normalization processing is performed, and the calculation formula is: The normalized data is represented by is the normalized data, respectively, the maximum value and the minimum value in the current data.​ 3.The personnel performance evaluation method based on IWOA-SVM according to claim 1, wherein, In step S3, the penalty factor c comes from the original form of SVM, and the kernel function parameter g comes from the SVM decision function; The SVM original form formula is: wherein, represents a slack variable, represents an interval allowing the deviation of the first data point, is a penalty factor; The IWOA-SVM model uses a SVM decision function, which introduces a Lagrange function on the basis of the original form of the SVM and a Gaussian kernel function wherein, is a Lagrange factor, is a Gaussian kernel function, is a kernel function parameter.

4. The personnel performance evaluation method based on IWOA-SVM according to claim 1, wherein, In step S3, the Tent mapping mode is: ； After generating chaotic value, the mapping rule of chaotic value and population space is: ； where: is the position of individual i in dimension j; and is the position boundary of individual i in dimension j; d is the dimension of the position.

5. The personnel performance evaluation method based on IWOA-SVM according to claim 1, wherein, In step S3, the pseudo-oppositional learning method is as follows: Let the individual whale... Location Individual whale The pseudo-opposing whale individual is Defined as: ； ； wherein: , is the opposite position of position , denotes a random value between p, q.

6. The personnel performance evaluation method based on IWOA-SVM according to claim 1, wherein, In the step S3, in the iterative calculation process of the position of the whale individual, the convergence factor a is a coefficient that decays from 2 to 0 as the iteration starts, and according to the current iteration number Suppose that the current iteration number is , , the maximum iteration number is n, then ; the behavior selection coefficient A and the random disturbance coefficient C are coefficients for controlling the behavior of the whale, and are set to , a random vector in [0, 1], then ; .

7. The personnel performance evaluation method based on IWOA-SVM according to claim 6, wherein, In step S3, in the iterative calculation process of the position of whale individual, the probability selection parameter p is a random number uniformly distributed in the range of [0, 1]; When and surrounding prey strategy is employed: ； ； wherein the position of the current best solution is denoted as , the position of a whale individual is denoted as , the distance between a whale individual and a prey is denoted as , is a step size; When time, and time, the spiral update strategy is adopted, and the whale individual and prey distance between ； ； wherein, is the distance of the whale individual to the prey, is a constant used to define the shape of the logarithmic spiral, the spiral shape parameter is a random number in [-1, 1]; In summary, when the overall mathematical model is described as follows: (1)。 8. The personnel performance evaluation method based on IWOA-SVM according to claim 7, wherein, In the step S3, when the whale updates its own position according to the randomly selected individual, instead of the non-optimal individual, and the update formula is: (2) (3) wherein the position vector of a randomly selected whale individual from the population is represented as the distance of the randomly selected whale individual to the prey is represented as .

9. The personnel performance evaluation method based on IWOA-SVM according to claim 8, wherein, In the step S3, in the iterative calculation process of the position of the whale individual, the differential evolution mechanism uses mutation expansion factor and crossover probability to effectively enhance the global search ability of the intelligent algorithm, specifically including mutation, crossover and greedy selection. ① Mutation operator is defined as follows: (4) In the formula: v is the mutation solution, x is the current solution, r1, r2, r3 represent random integers in [1, N], corresponding to three independent individuals, SF is the mutation expansion factor, which is used to control the change degree between the original individual and the mutation individual; ② Crossing operator, after individual mutation, the crossing operator can further enhance the population diversity on the target vector, and the differential evolution generates offspring individual zi on xi and vi by using the crossing operator, which is defined as: (5) In the formula: rand represents a random value in [0, 1], and CR represents the individual crossover probability; ③ Selection operator, the parent individual and its offspring individual are compared by using greedy strategy, and new individual selection is realized, which is defined as: (6) In the formula: f(z) is the fitness function value of z, which is defined as follows: ； Accuracy is the proportion of correctly classified samples to the total number of samples, as follows: ； In the formula: TP is true positive, FP is false positive, TN is true negative, and FN is false negative. The selection probability Pi is used to determine whether the whale individual updates the position according to the differential evolution strategy, and the selection probability of xi is defined as the ratio between the fitness of xi and the overall fitness of the population, which is expressed as: ； According to the selection probability Pi, the calculation method of the current solution xi is used to update the individual position, and the specific method is: If , rand is a random number in [0, 1], then the current solution xi is updated by equations (1)-(3); if , then the current solution xi is updated by equations (4)-(6).

Images