Cigarette business chain high-efficiency collaborative combined management evaluation method

Through cluster analysis, principal component analysis and Bi-GCN model, the redundancy of indicators and data missing in the cigarette industry's employee performance evaluation are solved, efficient employee behavior guidance and performance prediction are achieved, and management costs are reduced.

CN118941149APending Publication Date: 2024-11-12GUIYANG OFFICE OF GUIZHOU TOBACCO CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411011847.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing cigarette industry employee performance evaluation methods have problems such as redundant indicators, multiple guidance, and the problems of management confusion and high costs, as well as the lack of historical data and the difficulty in determining the changes in indicators over time.

Method used

Clustering analysis and principal component analysis are used to reduce index redundancy, time features are extracted through LSTM and employee work indicator values ​​are predicted using Bi-GCN model, and an undirected graph is constructed to calculate employee category distance, predict and evaluate employee performance.

Benefits of technology

It effectively reduces management chaos and high costs, solves the problems of employee behavior guidance obstacles and lack of performance data, and can dynamically capture changes in employee work indicators and determine work indicators.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941149A_ABST
    Figure CN118941149A_ABST
Patent Text Reader

Abstract

The invention discloses a cigarette business chain high-efficiency collaborative combined management evaluation method. The method comprises the following steps: acquiring historical data of employee work indexes, performing clustering analysis, screening important indexes by adopting principal component analysis, extracting time features through a long short-term memory network (LSTM), constructing an undirected graph, training by utilizing a bidirectional graph convolutional network (Bi-GCN), and finally evaluating the work index value of a target employee through a prediction model. According to the method, redundant indexes are reduced, and the problem that the working indexes are difficult to determine when the working index data of partial posts are missing and the indexes change along with time is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a high-efficiency collaborative joint management evaluation method for a cigarette business chain, and belongs to the field of information technology. Background Art

[0002] With the continuous development of the cigarette industry, how to effectively manage and evaluate employees' work performance has become a key issue in improving production efficiency and product quality.

[0003] The existing employee performance indicators are mainly obtained through uploading and collective discussion. Enterprises usually obtain a large number of comprehensive indicators. However, this also raises two problems: First, since the indicators come from various departments, they almost cover all production, operation and management fields of the enterprise. Therefore, there is often a strong correlation between the indicators. If these indicators are used directly to guide the behavior of departments and employees, there will be repeated management of the behavior of departments and employees caused by multiple guidance, which will lead to management confusion, high costs and behavioral guidance obstacles; second, through collective discussion, there are still some indicators that are difficult to reach a consensus on in some organizations, and whether to enter the database still requires scientific judgment; third, many positions have historical performance data missing and indicators change over time, and it is impossible to determine the work indicators. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a high-efficiency collaborative joint management evaluation method for a cigarette business chain to overcome the deficiencies of the prior art.

[0005] The technical solution of the present invention is: a method for evaluating the high-efficiency collaborative joint management of a cigarette business chain, the method comprising the following steps:

[0006] S01. Obtain historical data of work indicators of all employees;

[0007] S02. Perform cluster analysis on the employees based on the historical data of all employees' work indicators to obtain the employee category to which each employee belongs;

[0008] S03. Conduct principal component analysis on the indicators corresponding to each employee category;

[0009] S04. Select the principal component whose variance contribution rate is greater than the set threshold value according to the variance contribution rate of the principal component under each employee category;

[0010] S05, extracting time features from the historical data of employee work indicators of each employee category corresponding to the principal component through LSTM;

[0011] S06. Use time features to calculate the distance between employee categories to construct an undirected graph, and train a Bi-GCN prediction model through Bi-GCN based on the undirected graph;

[0012] S07. Predict the work index value of the target employee through the Bi-GCN prediction model;

[0013] S08. Use work indicator values ​​to evaluate target employees through a comprehensive evaluation model.

[0014] Furthermore, the method of performing principal component analysis on the indicators corresponding to each employee category is as follows: constructing an original data matrix X,

[0015]

[0016] Normalize the original matrix to the standardized matrix Z,

[0017]

[0018] in, i=1,2,...,m;j=1,2,...,n;m represents the number of groups of work indicators of employee categories used for principal component analysis, and n represents the maximum number of indicators contained in the work indicators of employee categories used for principal component analysis;

[0019] Perform SVD decomposition on the standardized matrix Z to obtain n singular value matrices Σ and corresponding left singular vectors U and right singular vectors V T ,

[0020] Z=UΣV T

[0021]

[0022] Among them, λ j Represents the singular values ​​of Z after SVD decomposition.

[0023] Furthermore, the calculation method of the variance contribution rate is:

[0024]

[0025] Furthermore, the step of selecting the principal component whose variance contribution rate is greater than a set threshold according to the variance contribution rate of the principal component under each employee category specifically includes:

[0026] Sort the variance contribution rates from large to small;

[0027] The sorted variance contribution rates are accumulated from large to small to obtain the cumulative variance contribution rate;

[0028] The singular value just before the singular value at which the cumulative variance contribution rate is greater than 85% and the corresponding left singular vector and right singular vector are selected to reconstruct the normalized matrix.

[0029] Furthermore, the employee work index historical data of each employee category corresponding to the principal component is obtained by

[0030] The method of LSTM to extract time features is:

[0031] The encoder and decoder in the Seq2seq model are set as LSTM units. The encoder updates the hidden state at each time point in chronological order, and then imports the hidden state at each time point into the decoder for decoding update. The attention mechanism is introduced in the decoder structure to assign different attention weights to the output vector in the hidden layer to capture the temporal feature information.

[0032] Furthermore, the method of using time features to calculate the distance between employee categories and construct an undirected graph is:

[0033] Calculate the distance between employee categories,

[0034]

[0035] Where r represents the rth employee category, s represents the sth employee category, and x rp represents the p-th time feature of the r-th employee category, x sp represents the p-th time feature of the s-th employee category;

[0036] Construct the distance matrix D,

[0037]

[0038] Consider each employee as a node in the graph;

[0039] Using the calculated distance matrix D, a distance threshold is set. When the distance between two nodes (employees) is less than the threshold, an undirected edge is established between the two nodes.

[0040] The beneficial effects of the present invention are as follows: compared with the prior art, the present invention reduces the redundancy between indicators through cluster analysis and principal component analysis, and avoids the management confusion and high cost problems caused by multiple guidance. By extracting time features through LSTM, the model can capture the dynamic changes of employee work indicators, and then use the Bi-GCN model to effectively predict the work indicators of employees with missing performance data, and the work indicators of employees with missing data also change dynamically. On the premise of solving the confusion and high cost of existing work indicators, the present invention also solves the problem that it is difficult to determine the work indicators when the work indicator data of some positions is missing and the indicators change over time. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0042] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0043] Implementation Example 1: Reference Figure 1 , a high-efficiency collaborative joint management evaluation method for a cigarette business chain, the method comprising the following steps:

[0044] S01. Obtain historical data of work indicators of all employees;

[0045] S02. Perform cluster analysis on the employees based on the historical data of all employees' work indicators to obtain the employee category to which each employee belongs;

[0046] S03. Conduct principal component analysis on the indicators corresponding to each employee category;

[0047] S04. Select the principal component whose variance contribution rate is greater than the set threshold value according to the variance contribution rate of the principal component under each employee category;

[0048] S05, extracting time features from the historical data of employee work indicators of each employee category corresponding to the principal component through LSTM;

[0049] S06. Use time features to calculate the distance between employee categories to construct an undirected graph, and train a Bi-GCN prediction model through Bi-GCN based on the undirected graph;

[0050] S07. Predict the work index value of the target employee through the Bi-GCN prediction model;

[0051] S08. Use work indicator values ​​to evaluate target employees through a comprehensive evaluation model.

[0052] Furthermore, the method of performing principal component analysis on the indicators corresponding to each employee category is:

[0053] Construct the original data matrix X,

[0054]

[0055] Normalize the original matrix to the standardized matrix Z,

[0056]

[0057] in, i=1,2,...,m;j=1,2,...,n;m represents the number of groups of work indicators of employee categories used for principal component analysis, and n represents the maximum number of indicators contained in the work indicators of employee categories used for principal component analysis;

[0058] Perform SVD decomposition on the standardized matrix Z to obtain n singular value matrices Σ and corresponding left singular vectors U and right singular vectors V T ,

[0059] Z=UΣV T

[0060]

[0061] Among them, λ j Represents the singular values ​​of Z after SVD decomposition.

[0062] Furthermore, the calculation method of the variance contribution rate is:

[0063]

[0064] Furthermore, the step of selecting the principal component whose variance contribution rate is greater than a set threshold according to the variance contribution rate of the principal component under each employee category specifically includes:

[0065] Sort the variance contribution rates from large to small;

[0066] The sorted variance contribution rates are accumulated from large to small to obtain the cumulative variance contribution rate;

[0067] The singular value just before the singular value at which the cumulative variance contribution rate is greater than 85% and the corresponding left singular vector and right singular vector are selected to reconstruct the normalized matrix.

[0068] Furthermore, the method of extracting time features from the employee work indicator history data of each employee category corresponding to the principal component through LSTM is:

[0069] The encoder and decoder in the Seq2seq model are set as LSTM units. The encoder updates the hidden state at each time point in chronological order, and then imports the hidden state at each time point into the decoder for decoding update. The attention mechanism is introduced in the decoder structure to assign different attention weights to the output vector in the hidden layer to capture the temporal feature information.

[0070] Furthermore, the method of using time features to calculate the distance between employee categories and constructing an undirected graph is: calculating the distance between employee categories,

[0071]

[0072] Where r represents the rth employee category, s represents the sth employee category, and x sp represents the p-th time feature of the r-th employee category, x sp represents the p-th time feature of the s-th employee category;

[0073] Construct the distance matrix D,

[0074]

[0075] Consider each employee as a node in the graph;

[0076] Using the calculated distance matrix D, a distance threshold is set. When the distance between two nodes (employees) is less than the threshold, an undirected edge is established between the two nodes.

[0077] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A high-efficiency collaborative joint management evaluation method for a cigarette business chain, characterized in that: The method comprises the following steps: S01. Obtain historical data of work indicators of all employees; S02. Perform cluster analysis on the employees based on the historical data of all employees' work indicators to obtain the employee category to which each employee belongs; S03. Conduct principal component analysis on the indicators corresponding to each employee category; S04. Select the principal component whose variance contribution rate is greater than the set threshold value according to the variance contribution rate of the principal component under each employee category; S05, extracting time features from the historical data of employee work indicators of each employee category corresponding to the principal component through LSTM; S06. Use time features to calculate the distance between employee categories to construct an undirected graph, and train a Bi-GCN prediction model through Bi-GCN based on the undirected graph; S07. Predict the work index value of the target employee through the Bi-GCN prediction model; S08. Use work indicator values ​​to evaluate target employees through a comprehensive evaluation model.

2. The high-efficiency collaborative joint management evaluation method for the cigarette business chain according to claim 1 is characterized in that: The method for performing principal component analysis on the indicators corresponding to each employee category is: Construct the original data matrix X, Normalize the original matrix to the standardized matrix Z, in, i=1,2,...,m;j=1,2,...,n;m represents the number of groups of work indicators of employee categories used for principal component analysis, and n represents the maximum number of indicators contained in the work indicators of employee categories used for principal component analysis; Perform SVD decomposition on the standardized matrix Z to obtain n singular value matrices Σ and corresponding left singular vectors U and right singular vectors V T , Z=UΣV T Among them, λ j Represents the singular values ​​of Z after SVD decomposition.

3. The high-efficiency collaborative joint management evaluation method for the cigarette business chain according to claim 2 is characterized in that: The calculation method of the variance contribution rate is:

4. The high-efficiency collaborative joint management evaluation method for the cigarette business chain according to claim 3 is characterized in that: The method of selecting the principal component whose variance contribution rate is greater than the set threshold according to the variance contribution rate of the principal component under each employee category specifically includes: Sort the variance contribution rates from large to small; The sorted variance contribution rates are accumulated from large to small to obtain the cumulative variance contribution rate; The singular value just before the singular value at which the cumulative variance contribution rate is greater than 85% and the corresponding left singular vector and right singular vector are selected to reconstruct the normalized matrix.

5. The high-efficiency collaborative joint management evaluation method for the cigarette business chain according to claim 3 is characterized in that: The method for extracting time features from the employee work indicator historical data of each employee category corresponding to the principal component through LSTM is: The encoder and decoder in the Seq2seq model are set as LSTM units. The encoder updates the hidden state at each time point in chronological order, and then imports the hidden state at each time point into the decoder for decoding update. The attention mechanism is introduced in the decoder structure to assign different attention weights to the output vector in the hidden layer to capture the temporal feature information.

6. The high-efficiency collaborative joint management evaluation method for the cigarette business chain according to claim 3 is characterized in that: The method of using time features to calculate the distance between employee categories and construct an undirected graph is: Calculate the distance between employee categories, Where r represents the rth employee category, s represents the sth employee category, and x rp represents the p-th time feature of the r-th employee category, x sp represents the p-th time feature of the s-th employee category; Construct the distance matrix D, Consider each employee as a node in the graph; Using the calculated distance matrix D, a distance threshold is set. When the distance between two nodes (employees) is less than the threshold, an undirected edge is established between the two nodes.