Employee assessment and incentive method based on integral conversion and machine learning

Abstract

The present application relates to the technical field of human resource informatization and data analysis, in particular to a kind of employee examination incentive method based on integral conversion and machine learning, including data acquisition and preprocessing, integral conversion based on weight, work performance examination based on random forest and clustering algorithm, performance salary association and optimization, result synchronization and feedback;Through built-in special integral conversion algorithm, the multidimensional accurate mapping of the core business workload of front-line skilled workers and points is realized, the objective examination of employee work performance is completed by combining the data analysis model based on machine learning, and the automatic association of points, examination results and performance salary is realized.

Description

Technical Field

[0001] This invention relates to the field of human resource information technology and data analysis technology, specifically to an employee performance evaluation and incentive method based on points conversion and machine learning. Background Technology

[0002] In the power industry, although preliminary information-based recording methods have been introduced for the performance evaluation of front-line skilled workers' core business workload, the following technical shortcomings exist:

[0003] The points conversion lacks algorithmic support: Current technology has not built a dedicated points conversion algorithm and quantification model. Points conversion relies solely on simple matching using preset fixed rules. The correspondence between employees' daily core business work content (such as work type, completion time, completion quality, work location, technical difficulty, labor intensity, etc.) and points lacks precise quantification supported by algorithms. The accuracy and scientific nature of points calculation cannot meet the needs of enterprises for refined and intelligent transformation of operation and management.

[0004] The performance evaluation system lacks data-driven approaches and is highly subjective: existing technologies do not incorporate data analysis models to deeply mine and assess employee work data. Performance evaluations rely on subjective human judgment, lacking algorithm-driven objective data support, and thus cannot accurately depict employees' actual work status, skill levels, and performance. This is detrimental to creating a fair competitive environment, hindering the activation of new employees and preventing employees from honing their skills on the front lines.

[0005] Inefficient and poorly adaptable performance-based compensation: Current technologies link performance evaluation results to performance-based pay through basic logic matching or manual operation, resulting in low efficiency and poor adaptability. This makes it difficult to quickly transform performance evaluation results into accurate and effective incentives. This hinders the development of a skills-based job advancement management system and fails to effectively address the technical challenges of employee motivation and skills enhancement in the context of hollowed-out frontline skills positions and a shortage of technical backbone personnel. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide an employee performance evaluation and incentive method based on points conversion and machine learning.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] An employee performance evaluation and incentive method based on points conversion and machine learning includes the following steps:

[0009] S1. Data Acquisition and Preprocessing

[0010] The system collects raw data on employee work through employee terminals, performs data preprocessing on the raw data, and transforms the various indicators in the raw data into multi-dimensional quantitative scores.

[0011] S2, Weight-based integral transformation

[0012] The weighting coefficients of each dimension's quantitative score are determined using the entropy method. Then, a weighted sum is calculated based on the quantitative score and the weighting coefficients to obtain the employee's daily score and monthly cumulative score.

[0013] S3. Performance evaluation based on random forest and clustering algorithms

[0014] Multi-dimensional behavioral features of employees during the assessment period are extracted, and assessment level labels are generated using the K-Means unsupervised clustering method. A random forest model is trained based on the multi-dimensional behavioral features and assessment level labels. The trained random forest model is used to evaluate the employees' monthly work performance and output an objective assessment level. The multi-dimensional behavioral features include: average daily points, point stability, task completion efficiency, quality pass rate, core business participation, and point growth trend.

[0015] S4. Performance-based Compensation Correlation and Optimization

[0016] Under the constraint of total salary budget, with the goal of maximizing the incentive effect of the employee's monthly points growth rate, the particle swarm optimization algorithm is used to solve the optimal performance-based salary mapping parameters, and the employee's monthly performance-based salary is calculated based on the optimal performance-based salary mapping parameters, the cumulative points for the month, and the assessment level.

[0017] S5. Results Synchronization and Feedback

[0018] The system stores employees' monthly accumulated points, assessment levels, and monthly performance-based pay in a cloud database and pushes them to employees' terminals simultaneously.

[0019] As a preferred embodiment, a further technical solution of the present invention is:

[0020] Preferably, the indicators in S1 include: technical difficulty, labor intensity, completion quality, completion efficiency, and safety standards;

[0021] Accordingly, the process of transforming the various indicators in the raw data into multi-dimensional quantitative scores includes:

[0022] (a) Technical Difficulty Score The determination

[0023] The technical difficulty score is obtained by looking up a table based on the job type:

[0024] ;

[0025] (ii) Labor intensity score The determination

[0026] Based on working hours Calculate the labor intensity score:

[0027] ;

[0028] (III) Completion Quality Score The determination

[0029] The completion quality score is obtained by mapping the quality evaluation level:

[0030] ;

[0031] (iv) Completion efficiency score The determination

[0032] The completion efficiency score is calculated based on the ratio of standard working hours to actual working hours.

[0033]

[0034] in, This indicates the preset standard working hours for the corresponding job type. Indicates the actual working hours completed by the employee. This indicates that the upper limit is set to 120% to prevent outliers from excessively affecting the integration results;

[0035] (v) Safety compliance score The determination

[0036] Safety compliance scores are derived by mapping safety performance:

[0037] .

[0038] Preferably, S2 specifically includes:

[0039] The points calculation for a single business transaction on that day is as follows:

[0040] ;

[0041] in, This represents the score for a single business task. , , , , This represents the weight coefficient for the corresponding dimension, satisfying... ;

[0042] Weighting coefficient Dynamically calculated using the entropy weight method, it can be expressed as:

[0043] ;

[0044] ;

[0045] ;

[0046] in, Indicates the first Entropy weight of the item indicator; Indicates the first Information entropy of the indicator, range of values ; This represents the total number of employees involved in the weighting calculation; Indicates the first The first item under the indicator The weighting of each employee's quantitative value; Indicates the first The employee in the The range of values ​​for each indicator after Min-Max standardization. ;

[0047] An employee's daily points are the sum of the points from all individual business tasks completed that day, and the monthly cumulative points are the sum of the employee's daily points for each workday.

[0048] Preferably, the average daily points in S3 Integral stability Task completion efficiency Quality pass rate Core business participation Points growth trend The calculations are as follows:

[0049] ;

[0050] in, Indicates the number of working days in the current month; This represents the employee's daily points on day t.

[0051] ;

[0052] in, This represents the standard deviation of an employee's daily points. This represents the average of the employees' points for the day;

[0053] ;

[0054] in, This indicates the total number of business tasks completed in the current month; Indicates the first Standard working hours for this business task; Indicates the first The actual working hours completed for each business task;

[0055] ;

[0056] in, This indicates the number of business tasks completed in the current month that met the standards of "qualified", "good", and "excellent". This indicates the number of business tasks completed in the current month;

[0057] ;

[0058] in, This indicates the number of business tasks that achieved a technical difficulty score of 80 or 100 in the current month;

[0059] ;

[0060] in, This represents the average daily points of employees for the current month. This represents the average daily points of employees in the previous month.

[0061] Preferably, the process of generating assessment grade labels using the K-Means unsupervised clustering method in S3 includes:

[0062] For each employee, a feature vector is constructed based on multi-dimensional behavioral characteristics: And construct a feature matrix based on the feature vectors of all employees;

[0063] For each feature in the feature matrix, Min-Max normalization is performed to map each feature to the [0,1] interval;

[0064] Set the number of clusters K=4, and use the K-Means++ method to initialize the cluster centers. Calculate the Euclidean distance from each employee to each cluster center using the feature vectors, and cluster the employees to the nearest cluster center according to the Euclidean distance, resulting in four clusters.

[0065] Calculate the mean of the corresponding elements in all features of each cluster, and use it as the new cluster center. Then return to execute the process of calculating the Euclidean distance from each employee to each cluster center using the feature vector, until the change in the cluster center is less than the preset threshold.

[0066] The cluster centers are sorted in descending order based on their daily average original scores, and then mapped to four assessment levels: Excellent, Good, Satisfactory, and Needs Improvement.

[0067] Preferably, the monthly performance-based compensation in S4 is calculated as follows:

[0068] ;

[0069] in, This represents the monthly performance-based pay for the i-th employee; This represents the base salary of the i-th employee; This represents the cumulative points for the i-th employee in the current month; This represents the highest cumulative score among all employees for the current month. Let i represent the quantitative value of the employee's performance evaluation level, with 1.0 for excellent, 0.8 for good, 0.6 for satisfactory, and 0.4 for needing improvement. These are the optimal performance-based compensation mapping parameters.

[0070] Preferably, the incentive effect of the employee's monthly points growth rate in S4 is quantified through the following steps:

[0071] Employee salary coefficient based on historical data Monthly accumulated points and the quantitative value of the assessment level The dataset is constructed using the actual monthly points growth rate as the input feature and the actual points growth rate of the following month as the training label.

[0072] The random forest regression model was trained using the dataset, and the monthly growth rate of employee points was predicted based on the trained random forest regression model.

[0073] Preferably, the process of using the particle swarm optimization algorithm to solve for the optimal performance-based pay mapping parameters in S4 includes:

[0074] Initialize the particle swarm, setting the number of particles in the swarm to 100, with each particle representing a set of candidate parameters. The initial positions of each particle are randomly generated within a set range, and the initial velocity is set to 0; the range of candidate parameter values ​​is set as follows: This represents the baseline value of the salary coefficient, which is the salary coefficient when both the cumulative points and the assessment level for the month are at their lowest. , represents the marginal contribution coefficient of the integral dimension to the salary coefficient, reflecting the degree of impact of changes in employee integrals on salary; This represents the marginal contribution coefficient of the performance evaluation grade to the salary coefficient, reflecting the degree of impact of changes in the performance evaluation grade on salary.

[0075] The fitness function is defined as the sum of the predicted monthly integral growth rates:

[0076] ;

[0077] in, This represents the monthly growth rate of the i-th employee's points, as predicted by the random forest regression model. Indicates the total compensation budget; This represents the penalty coefficient, used to constrain solutions that exceed the total compensation budget;

[0078] Calculate the fitness of each particle and iteratively update the velocity and position of each particle:

[0079] ;

[0080] ;

[0081] in, This represents the velocity of the i-th particle in the d-th dimension during the t-th iteration, where d=1,2,3 correspond to... ; This represents the position of the i-th particle in the d-th dimension during the t-th iteration; The inertia weight is represented, decreasing linearly from 0.8 to 0.4; This represents the individual learning factor, set to 1.5. This represents the social learning factor, set to 1.5. , Represents a random number uniformly distributed within the interval [0,1]. This represents the optimal position found by the i-th particle in history, i.e., the individual optimal position; This represents the optimal position found by the particle swarm in history, i.e., the global optimum;

[0082] When the maximum number of iterations is reached, the particle with the highest fitness is used as the optimal performance-based pay mapping parameter.

[0083] Preferably, the raw data includes at least: task number, task type, start time, end time, completion quality level, acceptance result, and operator.

[0084] The present invention, which adopts the above technical solution, has the following prominent features compared with the prior art:

[0085] This solution addresses the shortcomings of existing technologies. The lack of a dedicated point conversion algorithm and data analysis model results in insufficient accuracy in point calculation, a lack of objectivity in performance evaluation, and poor correlation between point results and performance-based compensation. This makes it difficult to meet the needs of refined and intelligent enterprise operations management transformation, as well as the technical challenges of precise point-based assessment and objective evaluation incentives required for skilled job advancement management systems. The solution utilizes algorithmic and model-supported information technology to achieve accurate conversion of employee work content into points, scientific assessment of work performance, and efficient correlation between point results and performance-based compensation. Attached Figure Description

[0086] Figure 1 This is a flowchart illustrating the employee performance evaluation and incentive method based on integral conversion and machine learning in an embodiment of the present invention. Detailed Implementation

[0087] The present invention will be further illustrated below with reference to specific embodiments. The purpose of this illustration is solely to provide a better understanding of the invention. Therefore, the examples given do not limit the scope of protection of the present invention.

[0088] like Figure 1 As shown in the figure, this embodiment presents an employee performance evaluation and incentive method based on integral conversion and machine learning, including the following steps:

[0089] S1. Data Acquisition and Preprocessing

[0090] The system collects raw data on employee work through employee terminals, preprocesses the raw data, and transforms various indicators in the raw data into multi-dimensional quantitative scores.

[0091] During implementation, the employee terminal can be a mobile phone, supporting touch input and photo upload functions. The original data of the employee's business work should include at least: task number, work type, start time, end time, completion quality level, acceptance result, and operator.

[0092] When collecting raw data, you can use a QR code scanning method: scan the QR code generated by the task work order to automatically import data such as job type, job duration, and job location; or you can use a manual input method: manually input data such as job quality, completion efficiency, and compliance with safety regulations through the terminal touch screen. The data collection terminal automatically records the data collection time, employee number, and terminal number to ensure data traceability. After collection, the data is uploaded to the State Grid cloud server in real time via the State Grid private network. The State Grid cloud service serves as the core computing and deployment platform, replacing traditional industrial-grade application servers. Leveraging the high-performance computing resources, elastic scalability, and security protection system of the State Grid cloud, it deploys the points conversion algorithm, data analysis model, and various software modules, undertaking the core roles of data computation, logical processing, and instruction scheduling. This ensures the efficiency, stability, and security of the algorithm model, meeting the real-time processing and dynamic expansion needs of large amounts of employee work data, while also conforming to the unified standards of the State Grid system's information technology construction. Furthermore, the State Grid cloud server's built-in relational cloud database (compatible with MySQL and Oracle) can replace traditional dedicated database servers, configuring elastic storage resources (which can be dynamically expanded according to data volume) to store all core data, including basic employee information, core business types and points standard data, daily employee work data, points conversion results, performance evaluation data, and performance-based compensation rules data. This provides secure and efficient data support for algorithm computation and model analysis, while also enabling automatic data backup, secure storage, and rapid retrieval via the State Grid cloud.

[0093] After collecting the raw data, proceed as follows:

[0094] Data Validation: The collected raw data is validated, and invalid data with format errors (such as job type not filled in, duration not being a number) or logical contradictions (such as job duration > 12 hours, job quality conflicting with safety regulations) is removed; abnormal data (such as job duration = 0.5 hours, job quality not selected) is marked and the administrator is reminded to verify it, and the verification record is synchronously stored in the State Grid Cloud Server log; if the verification fails, an error message is returned and the employee is guided to correct it. Its purpose is to realize the standardization and normalization of the collection of employees' daily core business work data, to provide real and complete raw data for subsequent points conversion and model analysis, and to solve the problem of inaccurate points conversion caused by non-standard and incomplete data collection in the existing technology.

[0095] Data cleaning: The original data is deduplicated and completed. For duplicate submissions (multiple submissions of the same job by the same employee), the latest one is retained. Missing non-critical auxiliary fields (such as job remarks) are filled with the default value "none" to ensure data uniqueness and integrity.

[0096] Data standardization: The raw data is standardized according to a preset format, the job type is converted into a unified code, the job duration is converted into minutes (e.g., 8 hours and 30 minutes is converted into 510 minutes), the job quality level (excellent, good, qualified) is quantified into 100 points, 80 points, and 60 points, and the completion efficiency is calculated as "actual completion time / standard time" into a quantitative score (e.g., standard time is 8 hours, actual completion time is 7 hours, efficiency score = 100 points), thus forming standardized data.

[0097] Specifically, the indicators include: technical difficulty, labor intensity, completion quality, completion efficiency, and safety standards;

[0098] Accordingly, the process of transforming the various indicators in the raw data into multi-dimensional quantitative scores includes:

[0099] (a) Technical Difficulty Score The determination

[0100] The technical difficulty score is obtained by looking up a table based on the job type:

[0101] ;

[0102] (ii) Labor intensity score The determination

[0103] Based on working hours Calculate the labor intensity score:

[0104] ;

[0105] (III) Completion Quality Score The determination

[0106] The completion quality score is obtained by mapping the quality evaluation level:

[0107] ;

[0108] (iv) Completion efficiency score The determination

[0109] The completion efficiency score is calculated based on the ratio of standard working hours to actual working hours.

[0110]

[0111] in, This indicates the preset standard working hours for the corresponding job type. Indicates the actual working hours completed by the employee. This indicates that the upper limit is set to 120% to prevent outliers from excessively affecting the integration results;

[0112] (v) Safety compliance score The determination

[0113] Safety compliance scores are derived by mapping safety performance:

[0114] .

[0115] S2, Weight-based integral transformation

[0116] The weighting coefficients of each dimension's quantitative score are determined using the entropy method. Then, a weighted sum is calculated based on the quantitative score and the weighting coefficients to obtain the employee's daily score and monthly cumulative score.

[0117] In practice, the points for a single business transaction on a given day are calculated as follows:

[0118] ;

[0119] in, This represents the score for a single business task. , , , , This represents the weight coefficient for the corresponding dimension, satisfying... ;

[0120] Weighting coefficient Dynamically calculated using the entropy weight method, it can be expressed as:

[0121] ;

[0122] ;

[0123] ;

[0124] in, Indicates the first Entropy weight of the item indicator; Indicates the first Information entropy of the indicator, range of values ; This represents the total number of employees involved in the weighting calculation; Indicates the first The first item under the indicator The weighting of each employee's quantitative value; Indicates the first The employee in the The range of values ​​for each indicator after Min-Max standardization. ;

[0125] An employee's daily points are the sum of points from all individual business tasks completed that day, and the monthly cumulative points are the sum of the employee's daily points from each workday. A dedicated algorithm is used to achieve a precise mapping between work content and points, solving the technical problems of insufficient algorithmic support and accuracy in existing point conversion methods, and ensuring the scientific and objective nature of point conversion.

[0126] Example: Weighting coefficients (task difficulty 0.3, task intensity 0.25, task quality 0.2, completion efficiency 0.15, safety compliance 0.1) are used to assign dimensional weights to the standardized data of each employee.

[0127] If employee A completes one task on a given day: "Equipment maintenance (medium difficulty, 80 points), 7 hours of work (80 points), excellent work quality (100 points), 100 points of completion efficiency, and full compliance with safety regulations (100 points)," then:

[0128] Technical difficulty score = 80 × 0.3 = 24 points; Labor intensity score = 80 × 0.25 = 20 points; Completion quality score = 100 × 0.2 = 20 points; Completion efficiency score = 100 × 0.15 = 15 points; Safety compliance score = 100 × 0.1 = 10 points;

[0129] Employee A's daily points = 24 + 20 + 20 + 15 + 10 = 89 points; at the same time, the point calculation process (scores, weights, and summary results for each dimension) is recorded to form detailed point data.

[0130] S3. Performance evaluation based on random forest and clustering algorithms

[0131] Multi-dimensional behavioral features of employees during the assessment period are extracted, and assessment level labels are generated using the K-Means unsupervised clustering method. A random forest model is trained based on the multi-dimensional behavioral features and assessment level labels. The trained random forest model is used to evaluate the employees' monthly work performance and output an objective assessment level. The multi-dimensional behavioral features include: average daily points, point stability, task completion efficiency, quality pass rate, core business participation, and point growth trend.

[0132] In practice, the average integral Integral stability Task completion efficiency Quality pass rate Core business participation Points growth trend The calculations are as follows:

[0133] ;

[0134] in, Indicates the number of working days in the current month; This represents the employee's daily points on day t.

[0135] ;

[0136] in, This represents the standard deviation of an employee's daily points. This represents the average of the employees' points for the day;

[0137] ;

[0138] in, This indicates the total number of business tasks completed in the current month; Indicates the first Standard working hours for this business task; Indicates the first The actual working hours completed for each business task;

[0139] ;

[0140] in, This indicates the number of business tasks completed in the current month that met the standards of "qualified", "good", and "excellent". This indicates the number of business tasks completed in the current month;

[0141] ;

[0142] in, This indicates the number of business tasks that achieved a technical difficulty score of 80 or 100 in the current month;

[0143] ;

[0144] in, This represents the average daily points of employees for the current month. This represents the average daily points of employees in the previous month.

[0145] In implementation, the process of generating assessment level labels using the K-Means unsupervised clustering method includes:

[0146] For each employee, a feature vector is constructed based on multi-dimensional behavioral characteristics: And construct a feature matrix based on the feature vectors of all employees;

[0147] For each feature in the feature matrix, Min-Max normalization is performed to map each feature to the [0,1] interval;

[0148] Set the number of clusters K=4, and use the K-Means++ method to initialize the cluster centers. Calculate the Euclidean distance from each employee to each cluster center using the feature vectors, and cluster the employees to the nearest cluster center according to the Euclidean distance, resulting in four clusters.

[0149] Calculate the mean of the corresponding elements in all features of each cluster, and use it as the new cluster center. Then return to execute the process of calculating the Euclidean distance from each employee to each cluster center using the feature vector, until the change in the cluster center is less than the preset threshold.

[0150] The cluster centers are sorted in descending order based on their daily average original scores, and then mapped to four assessment levels: Excellent, Good, Satisfactory, and Needs Improvement.

[0151] The random forest model is trained using multi-dimensional behavioral features as input features and K-Means clustering-generated level labels as training targets. The model parameters are set to 100 decision trees, 8 decision tree depths, and a minimum number of splits of 5. Warm_start=True is set so that when new data arrives each month, full retraining is not performed, but the partial_fit method is called to perform multiple rounds of incremental fitting. A sliding window mechanism is also used to retain only the data from the most recent 12 months for model updates and discard the data from the earliest month.

[0152] S4. Performance-based Compensation Correlation and Optimization

[0153] Under the constraint of the total salary budget, with the goal of maximizing the incentive effect of the employee's monthly points growth rate, the particle swarm optimization algorithm is used to solve the optimal performance-based salary mapping parameters, and the employee's monthly performance-based salary is calculated based on the optimal performance-based salary mapping parameters, the cumulative points for the month, and the assessment level.

[0154] In practice, the monthly performance-based compensation is calculated as follows:

[0155] ;

[0156] in, This represents the monthly performance-based pay for the i-th employee; This represents the base salary of the i-th employee; This represents the cumulative points for the i-th employee in the current month; This represents the highest cumulative score among all employees for the current month. Let i represent the quantitative value of the employee's performance evaluation level, with 1.0 for excellent, 0.8 for good, 0.6 for satisfactory, and 0.4 for needing improvement. These are the optimal performance-based compensation mapping parameters.

[0157] In practice, the incentive effect of the employee's monthly points growth rate is quantified through the following steps:

[0158] Employee salary coefficient based on historical data Monthly accumulated points and the quantitative value of the assessment level The dataset is constructed using the actual monthly points growth rate as the input feature and the actual points growth rate of the following month as the training label.

[0159] The random forest regression model was trained using the dataset, and the monthly growth rate of employee points was predicted based on the trained random forest regression model.

[0160] In practice, the process of using the particle swarm optimization algorithm to solve for the optimal performance-based pay mapping parameters includes:

[0161] Initialize the particle swarm, setting the number of particles in the swarm to 100, with each particle representing a set of candidate parameters. The initial positions of each particle are randomly generated within a set range, and the initial velocity is set to 0; the range of candidate parameter values ​​is set as follows: This represents the baseline value of the salary coefficient, which is the salary coefficient when both the cumulative points and the assessment level for the month are at their lowest. , represents the marginal contribution coefficient of the integral dimension to the salary coefficient, reflecting the degree of impact of changes in employee integrals on salary; This represents the marginal contribution coefficient of the performance evaluation grade to the salary coefficient, reflecting the degree of impact of changes in the performance evaluation grade on salary.

[0162] The fitness function is defined as the sum of the predicted monthly integral growth rates:

[0163] ;

[0164] in, This represents the monthly growth rate of the i-th employee's points, as predicted by the random forest regression model. Indicates the total compensation budget; This represents the penalty coefficient, used to constrain solutions that exceed the total compensation budget;

[0165] Calculate the fitness of each particle and iteratively update the velocity and position of each particle:

[0166] ;

[0167] ;

[0168] in, This represents the velocity of the i-th particle in the d-th dimension during the t-th iteration, where d=1,2,3 correspond to... ; This represents the position of the i-th particle in the d-th dimension during the t-th iteration; The inertia weight is represented, decreasing linearly from 0.8 to 0.4; This represents the individual learning factor, set to 1.5. This represents the social learning factor, set to 1.5. , Represents a random number uniformly distributed within the interval [0,1]. This represents the optimal position found by the i-th particle in history, i.e., the individual optimal position; This represents the optimal position found by the particle swarm in history, i.e., the global optimum;

[0169] When the maximum number of iterations is reached, the particle with the highest fitness is used as the optimal performance-based pay mapping parameter.

[0170] S5. Results Synchronization and Feedback

[0171] The system stores employees' monthly accumulated points, assessment levels, and monthly performance-based pay in a cloud database and pushes them to employees' terminals simultaneously.

[0172] During implementation, operation logs can also be generated, including operation records (administrator parameter adjustments, employee data submissions), fault records (abnormality type, processing process, processing result), data processing records (point conversion, data analysis process), and parameter adjustment records. The logs are retained for one year and stored on the State Grid cloud server. Administrators can remotely export the logs through the terminal for troubleshooting and system optimization. The log management complies with the State Grid log retention standards.

[0173] This invention constructs a dedicated points conversion algorithm to transform abstract indicators such as the technical difficulty and labor intensity of core businesses into quantifiable weight parameters. Combined with the actual completion of employees' work (duration, quality, etc.), it achieves accurate mapping between work content and points through multi-dimensional weighted calculation, solving the problems of insufficient algorithm support and accuracy in points conversion in existing technologies, and ensuring the scientific nature of points calculation.

[0174] A data analysis model is built based on machine learning algorithms. By deeply mining employees' historical score data and original work data, the core features reflecting employees' work performance are extracted, realizing data-driven objective evaluation, avoiding the subjectivity of manual evaluation, and solving the problem of lack of objective support in the performance evaluation in the existing technology.

[0175] By using a performance-based pay association algorithm, the association rules between points, evaluation results and pay are preset to achieve automatic matching and calculation of the three, replacing the traditional manual association operation, improving the efficiency and accuracy of association, and solving the problems of poor association adaptability and low efficiency in existing technologies.

[0176] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the invention. All equivalent changes made based on the description and drawings of the present invention are included within the scope of the present invention.

Claims

1. An employee performance evaluation and incentive method based on integral conversion and machine learning, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing The system collects raw data on employee work through employee terminals, performs data preprocessing on the raw data, and transforms the various indicators in the raw data into multi-dimensional quantitative scores. S2, Weight-based integral transformation The weighting coefficients of each dimension's quantitative score are determined using the entropy method. Then, a weighted sum is calculated based on the quantitative score and the weighting coefficients to obtain the employee's daily score and monthly cumulative score. S3. Performance evaluation based on random forest and clustering algorithms Multi-dimensional behavioral features of employees during the assessment period are extracted, and assessment level labels are generated using the K-Means unsupervised clustering method. A random forest model is trained based on the multi-dimensional behavioral features and assessment level labels. The trained random forest model is used to evaluate the employees' monthly work performance and output an objective assessment level. The multi-dimensional behavioral features include: average daily points, point stability, task completion efficiency, quality pass rate, core business participation, and point growth trend. S4. Performance-based Compensation Correlation and Optimization Under the constraint of total salary budget, with the goal of maximizing the incentive effect of the employee's monthly points growth rate, the particle swarm optimization algorithm is used to solve the optimal performance-based salary mapping parameters, and the employee's monthly performance-based salary is calculated based on the optimal performance-based salary mapping parameters, the cumulative points for the month, and the assessment level. S5. Results Synchronization and Feedback The system stores employees' monthly accumulated points, assessment levels, and monthly performance-based pay in a cloud database and pushes them to employees' terminals simultaneously.

2. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 1, characterized in that, The indicators in S1 include: technical difficulty, labor intensity, completion quality, completion efficiency, and safety standards; Accordingly, the process of transforming the various indicators in the raw data into multi-dimensional quantitative scores includes: (a) Technical Difficulty Score The determination The technical difficulty score is obtained by looking up a table based on the job type: ； (ii) Labor intensity score The determination Based on working hours Calculate the labor intensity score: ； (III) Completion Quality Score The determination The completion quality score is obtained by mapping the quality evaluation level: ； (iv) Completion efficiency score The determination The completion efficiency score is calculated based on the ratio of standard working hours to actual working hours. ； in, This indicates the preset standard working hours for the corresponding job type. Indicates the actual working hours completed by the employee. This indicates that the upper limit is set to 120% to prevent outliers from excessively affecting the integration results; (v) Safety compliance score The determination Safety compliance scores are derived by mapping safety performance: 。 3. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 2, characterized in that, S2 specifically includes: The points calculation for a single business transaction on that day is as follows: ； in, This represents the score for a single business task. , , , , This represents the weight coefficient for the corresponding dimension, satisfying... ; Weighting coefficient Dynamically calculated using the entropy weight method, it can be expressed as: ； ； ； in, Indicates the first Entropy weight of the item indicator; Indicates the first Information entropy of the indicator, range of values ; This represents the total number of employees involved in the weighting calculation; Indicates the first The first item under the indicator The weighting of each employee's quantitative value; Indicates the first The employee in the The range of values ​​for each indicator after Min-Max standardization. ; An employee's daily points are the sum of the points from all individual business tasks completed that day, and the monthly cumulative points are the sum of the employee's daily points for each workday.

4. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 3, characterized in that, S3 Daily Average Points Integral stability Task completion efficiency Quality pass rate Core business participation Points growth trend The calculations are as follows: ； in, Indicates the number of working days in the current month; This represents the employee's daily points on day t. ； in, This represents the standard deviation of an employee's daily points. This represents the average of the employee's points for the day; ； in, This indicates the total number of business tasks completed in the current month; Indicates the first Standard working hours for this business task; Indicates the first The actual working hours completed for each business task; ； in, This indicates the number of business tasks completed in the current month that met the standards of "qualified", "good", and "excellent". This indicates the number of business tasks completed in the current month; ； in, This indicates the number of business tasks that achieved a technical difficulty score of 80 or 100 in the current month; ； in, This represents the average daily points of employees for the current month. This represents the average daily points of employees in the previous month.

5. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 4, characterized in that, The process of generating assessment grade labels using the K-Means unsupervised clustering method in S3 includes: For each employee, a feature vector is constructed based on multi-dimensional behavioral characteristics: And construct a feature matrix based on the feature vectors of all employees; For each feature in the feature matrix, Min-Max normalization is performed to map each feature to the [0,1] interval; Set the number of clusters K=4, and use the K-Means++ method to initialize the cluster centers. Calculate the Euclidean distance from each employee to each cluster center using the feature vectors, and cluster the employees to the nearest cluster center according to the Euclidean distance, resulting in four clusters; Calculate the mean of the corresponding elements in all features in each cluster, and use it as the new cluster center. Then return to execute the process of calculating the Euclidean distance from each employee to each cluster center using the feature vector, until the change in the cluster center is less than the preset threshold. The cluster centers are sorted in descending order based on their daily average original scores, and then mapped to four assessment levels: Excellent, Good, Satisfactory, and Needs Improvement.

6. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 5, characterized in that, The monthly performance-based compensation calculation in S4 is as follows: ； in, This represents the monthly performance-based pay for the i-th employee; This represents the base salary of the i-th employee; This represents the cumulative points for the i-th employee in the current month; This represents the highest cumulative score among all employees for the current month. Let i represent the quantitative value of the employee's performance evaluation level, with 1.0 for excellent, 0.8 for good, 0.6 for satisfactory, and 0.4 for needing improvement. These are the optimal performance-based compensation mapping parameters.

7. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 6, characterized in that, The incentive effect of employee points growth rate in S4 is quantified through the following steps: Employee salary coefficient based on historical data Monthly accumulated points and the quantitative value of the assessment level The dataset is constructed using the actual monthly points growth rate as the input feature and the actual points growth rate of the following month as the training label. The random forest regression model was trained using the dataset, and the monthly growth rate of employee points was predicted based on the trained random forest regression model.

8. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 7, characterized in that, The process of using the particle swarm optimization algorithm in S4 to solve for the optimal performance-based pay mapping parameters includes: Initialize the particle swarm, setting the number of particles in the swarm to 100, with each particle representing a set of candidate parameters. The initial positions of each particle are randomly generated within a set range, and the initial velocity is set to 0; the range of candidate parameter values ​​is set as follows: This represents the baseline value of the salary coefficient, which is the salary coefficient when both the cumulative points and the assessment level for the month are at their lowest. , represents the marginal contribution coefficient of the integral dimension to the salary coefficient, reflecting the degree of impact of changes in employee integrals on salary; This represents the marginal contribution coefficient of the performance evaluation grade to the salary coefficient, reflecting the degree of impact of changes in the performance evaluation grade on salary. The fitness function is defined as the sum of the predicted monthly integral growth rates: ； in, This represents the monthly growth rate of the i-th employee's points, as predicted by the random forest regression model. Indicates the total compensation budget; This represents the penalty coefficient, used to constrain solutions that exceed the total compensation budget; Calculate the fitness of each particle and iteratively update the velocity and position of each particle: ； ； in, This represents the velocity of the i-th particle in the d-th dimension during the t-th iteration, where d=1,2,3 correspond to... ; This represents the position of the i-th particle in the d-th dimension during the t-th iteration; The inertia weight is represented, decreasing linearly from 0.8 to 0.4; This represents the individual learning factor, set to 1.

5. This represents the social learning factor, set to 1.

5. , Represents a random number uniformly distributed within the interval [0,1]. This represents the optimal position found by the i-th particle in history, i.e., the individual optimal position; This represents the optimal position found by the particle swarm in history, i.e., the global optimum; When the maximum number of iterations is reached, the particle with the highest fitness is used as the optimal performance-based pay mapping parameter.

9. The employee performance evaluation and incentive method based on integral conversion and machine learning according to claim 1, characterized in that, The raw data should include at least: job task number, job type, start time, end time, completion quality level, acceptance result, and operator.

Images