A method for constructing and clustering user portraits based on multi-dimensional attributes

Abstract

The present invention discloses a method for constructing and clustering user portraits based on multi-dimensional attributes, which relates to the technical field of user portrait construction and clustering. The method includes: Step 1: Obtain the production quantity, defective quantity and personnel files of workers in their historical work; Step 2: Analyze and obtain the work efficiency, skills, habits and stability dimensions of each worker according to the production quantity, defective quantity and personnel files of the workers in their historical work; Step 3: After normalizing the work efficiency, skills, habits and stability dimension values of each worker, construct the user portraits of each worker. Through the clustering method based on the similarity value of node vectors, it is possible to accurately discover groups of workers with similar characteristics, help enterprises understand the work performance and characteristics of workers more comprehensively and deeply, provide a scientific basis for human resource management, be able to accurately classify workers, contribute to targeted training, motivation and job arrangement of enterprises, and improve the utilization efficiency of human resources.

Description

Technical Field

[0001] The present invention relates to the technical field of user portrait construction and clustering, and particularly relates to a method for constructing and clustering user portraits based on multi-dimensional attributes. Background Art

[0002] The application of the method for constructing and clustering user portraits based on multi-dimensional attributes to assembly line workers can help enterprises more comprehensively understand various characteristics of assembly line workers, so as to carry out more effective management and resource allocation;

[0003] However, in the human resource management of modern enterprises, with the expansion of enterprise scale and the complexity and diversification of business, more refined and scientific methods are required for the evaluation and management of workers. Traditional worker evaluation methods may be relatively single and subjective, unable to comprehensively and accurately reflect the true work performance and ability characteristics of workers. Enterprises lack analysis based on multi-dimensional data in human resource allocation and management, resulting in unreasonable personnel arrangements, affecting production efficiency and quality, and it is difficult to discover groups of workers with similar characteristics, and it is impossible to formulate targeted training, incentive and management strategies; based on this, a method for constructing and clustering user portraits based on multi-dimensional attributes is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for constructing and clustering user portraits based on multi-dimensional attributes, which solves the technical problem of AAAAAAA.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A method for constructing and clustering user portraits based on multi-dimensional attributes includes the following steps:

[0007] Step 1: Obtain the output quantity, defective quantity and personnel files in the historical work of workers;

[0008] Step 2: Analyze and obtain the work efficiency, skills, habits and stability dimensions of each worker according to the historical work output, defective quantity and personnel files of the workers;

[0009] Step 3: After normalizing the work efficiency, skills, habits and stability dimension values of each worker, represent them as a four-dimensional vector, and then construct the user portrait of each worker;

[0010] Step 4: Regard the workers as nodes and obtain node vectors at the same time, calculate the similarity values between the nodes, construct connections according to the calculation results of the similarity values, and then generate a connection diagram corresponding to each employee;

[0011] Step Five: Conduct a clustering analysis on the number of construction connection lines in the construction connection diagrams of each employee, group the user portraits with the same number of construction connection lines into one category, thereby obtaining different clusters of worker user portraits, and output and display the clustering results.

[0012] As a further solution of the present invention: The specific way to obtain the work efficiency dimension is as follows:

[0013] Obtain the production quantity and defective quantity of each worker within the working hours t, then mark the ratio between the defective quantity and the production quantity of each worker as the defective rate corresponding to each worker respectively, take the ratio between the production quantity of each worker within the working hours t and the working hours t as the production efficiency corresponding to each worker respectively, and mark the sum of the products of the production efficiency and the defective rate corresponding to each worker and the preset coefficients β1 and β2 respectively as the work efficiency dimension WAa of each worker, where β1 and β2 are both preset coefficients, satisfying 1 = β1 + β2, and β1 > β, t is in hours, a represents different workers, and at the same time a is the worker label corresponding to each worker respectively, and a is a positive integer.

[0014] As a further solution of the present invention: The specific way to obtain the work skill dimension is as follows:

[0015] Obtain the working years LEa and on-the-job years LFa corresponding to each worker from the personnel files, and obtain the work skill dimension WBa corresponding to each worker through WBa = LFa × β3 + (LFa / LEa) × β4, where β3 and β4 are both preset coefficients, satisfying 1 = β3 + β4, and β3 > β4.

[0016] As a further solution of the present invention: The specific way to obtain the work habit dimension is as follows:

[0017] Obtain the number of late arrivals corresponding to each worker in the preset number of days T from the attendance records, and mark the reciprocal of the ratio between the number of late arrivals and the preset number of days T as the work habit dimension WCa corresponding to each worker respectively. T is the preset number of days, and the specific value range is greater than 1 day and less than 90 days, and at the same time T is a positive integer.

[0018] As a further solution of the present invention: The specific way to obtain the work stability dimension is as follows:

[0019] Obtain the number of job-hopping times of each worker during their working years from the personnel files, and take the reciprocal of the ratio between the number of job-hopping times of each worker and their corresponding working years as the work stability dimension WDa of each worker.

[0020] As a further solution of the present invention: The specific construction method for constructing the user portrait of each worker is as follows;

[0021] Normalize the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension corresponding to each worker so that the values of each dimension are between 0 and 1. Use the normalized dimension values as the feature vector of the user portrait of the worker. That is, the user portrait of each worker can be represented as a four-dimensional vector (WAa, WBa, WCa, WDa); form a matrix with the user portraits of all workers. Each row represents the user portrait of a worker, and the columns correspond to the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension respectively, thereby constructing the user portrait of each worker.

[0022] As a further solution of the present invention: The specific method for generating the construction connection diagrams corresponding to each employee is as follows:

[0023] S1: Regard each worker as a different node. Arbitrarily select one worker from all workers as the point of the target worker as the calibration node P. Sort the other remaining nodes in ascending order according to the worker number a, and label them as Ub in turn. b represents other different remaining nodes, and at the same time, use b as the node number of other different remaining nodes. b is a positive integer;

[0024] S2: According to the four-dimensional vector of each worker's user portrait, obtain the node vector of the calibration node, denoted as P(DA1, DB1, DC1, DD1), and the node vectors of other different remaining nodes, denoted as Ub(EAb, EBb, ECb, EDb). Calculate the similarity value Hb between the node vector of the calibration node and the node vectors of other different remaining nodes through the similarity value calculation formula. S3: Construct a connection between the node with the similarity value Hb greater than 0.9 and the calibration node, thereby generating the construction connection diagram corresponding to the calibration node. Otherwise, do not perform any processing, and use it as the construction connection diagram of the employee corresponding to the calibration node;

[0025] S4: Repeat the above steps S1 - S3 to obtain the construction connection diagrams corresponding to each employee.

[0026] As a further solution of the present invention: The specific method for obtaining different clusters of worker user portraits is as follows:

[0027] Obtain the number of construction connections corresponding to each construction connection diagram of each employee. Group the workers with the same number of construction connections into the same cluster, and group the user portraits with different numbers of connections into a separate cluster, thereby obtaining different clusters of worker user portraits and outputting the clustering results at the same time.

[0028] The beneficial effects of the present invention:

[0029] (1) In the present invention, by effectively integrating the historical work records and personnel files of each worker on the assembly line, the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension corresponding to each worker are obtained. Furthermore, a user portrait of each worker is constructed, achieving a comprehensive analysis of the multi-dimensional attributes of each worker, comprehensively evaluating the worker's performance, avoiding the limitation of evaluating workers from a single dimension, but comprehensively considering multiple dimensions such as work efficiency, work skill, work habit, and work stability, more accurately reflecting the actual performance of workers, constructing a clear user portrait, and providing a reliable data basis for subsequent clustering analysis;

[0030] (2) In the present invention, through the clustering method based on the similarity value of node vectors, it is possible to accurately discover groups of workers with similar characteristics, helping the enterprise to more comprehensively and deeply understand the work performance and characteristics of workers, providing a scientific basis for human resource management, being able to accurately classify workers, contributing to the enterprise's targeted training, motivation, and job arrangement, improving the utilization efficiency of human resources. By considering dimensions such as work habit and stability, it helps to create a good working atmosphere, improve the overall work efficiency and quality. The clustering analysis based on multi-dimensional attributes can discover potential problems and trends, providing valuable references for the enterprise's strategic decision-making and management optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The present invention will be further described below with reference to the accompanying drawings.

[0032] Figure 1 It is a schematic diagram of the framework structure of a method for constructing a user portrait and clustering based on multi-dimensional attributes of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0034] Please refer to Figure 1 As shown, the present invention is a method for constructing a user portrait and clustering based on multi-dimensional attributes, including the following steps:

[0035] Step 1: Obtain the historical work records and personnel files of each worker. The historical work records include the quantity of production and the quantity of defective products;

[0036] The historical work records of workers are specifically collected and recorded by installing sensors and monitoring devices on the production line. The personnel files of workers are obtained from the personnel database, which are all existing and mature technologies, so no further elaboration will be made here.

[0037] Step 2: Analyze the historical work records and personnel files of each worker on the assembly line to obtain the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension corresponding to each worker.

[0038] The specific method for obtaining the work efficiency dimension is as follows:

[0039] From the personnel files, obtain the output quantity LAa of each worker within the working hours t, and at the same time obtain the number of defective products LBa of each worker within the working hours t. Then, mark the ratio between the number of defective products LBa and the output quantity LAa of each worker as the defective rate LCa corresponding to each worker, and use the ratio between the output quantity LAa of each worker within the working hours t and the working hours t as the production efficiency LDa corresponding to each worker, where a represents different workers, and at the same time a is the worker label corresponding to each worker, and a is a positive integer.

[0040] It should be noted that the working hours t is a preset positive integer in hours. Here, the specific value of t is 40 hours.

[0041] According to the production efficiency LDa and defective rate LCa corresponding to each worker, obtain the work efficiency dimension WAa corresponding to each worker through WAa = LDa×β1 + LCa×β2, where β1 and β2 are both preset coefficients, satisfying 1 = β1 + β2, and β1 > β2.

[0042] The specific method for obtaining the work skill dimension is as follows:

[0043] Obtain the working years LEa and on-the-job hours LFa corresponding to each worker, and obtain the work skill dimension WBa corresponding to each worker through WBa = LFa×β3 + (LFa / LEa)×β4, where β3 and β4 are both preset coefficients, satisfying 1 = β3 + β4, and β3 > β4.

[0044] It should be noted that both the working years and on-the-job hours can be obtained from the personnel files of workers. The working years refer to the length of time engaged in the corresponding work, and the on-the-job hours refer to the length of time working on this post.

[0045] The specific method for obtaining the work habit dimension is as follows:

[0046] Obtain the number of latenesses corresponding to each worker in the preset number of days T from the attendance records, and mark the reciprocal of the ratio between the number of latenesses and the preset number of days T as the work habit dimension WCa corresponding to each worker. T is the preset number of days, and the specific value range is greater than 1 day and less than 90 days. At the same time, T is a positive integer:

[0047] The specific method for obtaining the work stability dimension is as follows:

[0048] From the personnel files, obtain the number of job hops of each worker during their employment duration, and use the reciprocal of the ratio between the number of job hops of each worker and their corresponding employment duration as the work stability dimension WDa of each worker;

[0049] Step 3: Based on the work efficiency dimension WAa, work skill dimension WBa, work habit dimension WCa, and work stability dimension WDa corresponding to each worker obtained from the above analysis, construct a user portrait for each worker. The specific construction method is as follows:

[0050] First, perform normalization processing on the work efficiency dimension LEa, work skill dimension WBa, work habit dimension WCa, and work stability dimension WDa corresponding to each worker, so that the values of each dimension are between 0 and 1;

[0051] Then, use the normalized values of each dimension as the feature vectors of the user portrait of this worker. That is, the user portrait of each worker can be represented as a four-dimensional vector (WAa, WBa, WCa, WDa);

[0052] The normalization here is an existing and mature technology, so no further elaboration is made;

[0053] Finally, form a matrix with the user portraits of all workers. Each row represents the user portrait of a worker, and the columns correspond to the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension respectively, thereby constructing the user portrait of each worker;

[0054] Step 4: Consider each worker as a different node, and at the same time obtain the node vectors of each node according to the four-dimensional vectors of the user portraits of each worker. Calculate the similarity values between the node vectors of each node and the node vectors of the other remaining nodes, and construct connections based on the calculation results of the similarity values, thereby generating the construction connection diagrams corresponding to each employee. The specific method is as follows:

[0055] S1: Treat each worker as a different node. Arbitrarily select one worker from all the workers as the target worker, and use the node corresponding to the target worker as the calibration node. Mark the calibration node as P, sort the other remaining nodes in ascending order according to the worker label a, and sequentially label them as Ub, where b represents other different remaining nodes. At the same time, use b as the node label of other different remaining nodes, and b is a positive integer;

[0056] S2: According to the four-dimensional vector (WAa, WBa, WCa, WDa) of each worker's user portrait, obtain the four-dimensional vectors of each worker's user portrait as the node vectors of each node. Furthermore, mark the node vector of the calibration node as P(DA1, DB1, DC1, DD1), and mark the node vectors of other different remaining nodes as Ub(EAb, EBb, ECb, EDb);

[0057] Calculate the similarity value Hb between the node vector of the calibration node and the node vectors of other different remaining nodes through the similarity value calculation formula;

[0058] The specific similarity value calculation formula is:

[0059] Among them,

[0060] The closer the similarity value Hb is to 1, the more similar the worker portraits between the corresponding nodes and the calibration node are. Conversely, the less similar they are;

[0061] S3: Connect the nodes with similarity value Hb greater than 0.9 to the calibration node, and then generate the constructed connection graph corresponding to the calibration node. Conversely, do not perform any processing, and use it as the constructed connection graph of the employee corresponding to the calibration node;

[0062] S4: Repeat the above steps S1 - S3 to obtain the constructed connection graphs corresponding to each employee respectively;

[0063] Step Five: Perform clustering analysis on the user portraits of each worker according to the number of constructed connections corresponding to the constructed connection graphs corresponding to each employee respectively. Obtain different clusters of worker user portraits according to the analysis results. The specific method is:

[0064] Obtain the number of construction connection lines corresponding to each employee in the constructed connection diagram. This number reflects the similarity degree of each worker with other workers in the user portrait, that is, how many other workers' user portraits have a high enough similarity with this worker's user portrait, that is, the similarity value is greater than 0.9. Workers with the same number of construction connection lines are grouped into the same cluster. This means that if two or more workers have the same number of connection lines in different constructed connection diagrams, then these workers are considered to have similar characteristics in the user portrait and are thus classified into the same cluster. User portraits with different numbers of connection lines are grouped into a separate cluster. Then, different clusters of workers' user portraits are obtained, and at the same time, the clustering results are output. The clustering results are displayed in the form of tables, graphs, or other visualizations so that managers or decision-makers can intuitively understand the characteristics of workers in different clusters, and then take corresponding management measures or formulate targeted training plans;

[0065] For example, assume there are four workers A, B, C, and D. After the processing of steps one to four, we obtain the constructed connection diagrams of the four workers A, B, C, and D respectively. There are 2 connection lines in the constructed connection diagram of worker A, indicating that there are 2 workers' user portraits with a similarity greater than 0.9 to A. There are 3 connection lines in the constructed connection diagram of worker B, indicating that there are 3 workers' user portraits with a similarity greater than 0.9 to B. There are 2 connection lines in the constructed connection diagram of worker C, also indicating that there are 2 workers' user portraits with a similarity greater than 0.9 to C. There is 1 connection line in the constructed connection diagram of worker D, indicating that there is 1 worker's user portrait with a similarity greater than 0.9 to D. In this example, workers A and C have the same number of construction connection lines, both being 2, so they are grouped into the same cluster. Workers B and D are in separate clusters because their numbers of construction connection lines are different, being 3 and 1 respectively;

[0066] Finally, we obtain three clusters: Cluster 1 contains workers A and C, Cluster 2 contains worker B, and Cluster 3 contains worker D. Such clustering results help managers understand the characteristics of different types of workers, so as to adopt more precise management strategies;

[0067] Perform clustering analysis on the constructed user portraits of each worker to find groups of workers with similar characteristics. This method constructs the user portraits of workers through the analysis of multi-dimensional attributes and uses clustering analysis to find groups of workers with similar characteristics;

[0068] Through the analysis of these multi-dimensional attributes, the portrait of each worker can be constructed, and workers with similar attributes can be clustered, so as to facilitate the enterprise to carry out more targeted personnel management, training arrangement, job allocation, etc. Work, and cluster workers with similar attributes to facilitate the enterprise to carry out more targeted personnel management, training arrangement, job allocation, etc.;

[0069] Effectively integrates the historical work records and personnel files of each worker on the assembly line, obtains the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension corresponding to each worker respectively, and then constructs a user portrait for each worker, achieving a comprehensive analysis of the multi-dimensional attributes of each worker, comprehensively evaluating the worker performance, avoiding the limitation of evaluating workers from a single dimension, but comprehensively considering multiple dimensions such as work efficiency, work skill, work habit, and work stability, more accurately reflecting the actual performance of workers, constructing a clear user portrait, and providing a reliable data basis for subsequent clustering analysis;

[0070] Through the clustering method based on the similarity value of node vectors, it can accurately discover groups of workers with similar characteristics, help the enterprise understand the work performance and characteristics of workers more comprehensively and deeply, provide a scientific basis for human resource management, can accurately classify workers, contribute to the enterprise's targeted training, motivation, and job arrangement, improve the utilization efficiency of human resources, and through the consideration of dimensions such as work habit and stability, help create a good work atmosphere, improve the overall work efficiency and quality. The clustering analysis based on multi-dimensional attributes can discover potential problems and trends, providing valuable references for the enterprise's strategic decision-making and management optimization.

[0071] The above formulas are all dimensionless and take their numerical calculations. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and threshold selection in the formula are set by those skilled in the art according to the actual situation.

[0072] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claimed rights.

Claims

1. A user portrait construction and clustering method based on multi-dimensional attributes, characterized in that: The following steps are involved: Step 1: Obtain the worker's historical work output quantity and defective quantity as well as personnel files; Step 2: Based on the workers’ historical work output, defective product quantity and personnel files, analyze and obtain the work efficiency, skills, habits and stability dimensions of each worker; Step 3: Normalize the work efficiency, skills, habits, and stability dimension values ​​of each worker and express them as a four-dimensional vector, and then construct a user profile for each worker; Step 4: Treat the workers as nodes and obtain the vectors of each node at the same time, calculate the similarity values ​​between the nodes, build connections based on the similarity value calculation results, and then generate a connection diagram corresponding to each employee; Step 5: Perform cluster analysis on the number of construction links in the construction link diagram of each employee, classify user portraits with the same number of construction links into one category, and then obtain different clusters of worker user portraits, output the clustering results and display them; The specific method for obtaining the work efficiency dimension is: Obtain the output quantity and defective quantity of each worker within the working time t, then mark the ratio between the defective quantity and the output quantity of each worker as the defective rate corresponding to each worker, take the ratio between the output quantity and the working time t of each worker as the output efficiency corresponding to each worker, and mark the sum of the products of the output efficiency and defective rate corresponding to each worker with the preset coefficients β1 and β2 as the work efficiency dimension WAa ​​corresponding to each worker, where β1 and β2 are both preset coefficients, satisfying 1=β1+β2, and β1>β2, t is hours, a refers to different workers, and a is the worker number corresponding to each worker, and a is a positive integer; The specific way to obtain the work skill dimension is: Obtain the corresponding working hours LEa and working hours LFa of each worker from the personnel file, and obtain the corresponding work skill dimension WBa of each worker through WBa=LFa×β3+(LFa / LEa)×β4, where β3 and β4 are both preset coefficients, satisfying 1=β3+β4, and β3>β4; The specific method of obtaining the work habit dimension is: Obtain the number of latenesses of each worker in the preset days T from the attendance records, and mark the inverse of the ratio between the number of latenesses and the preset days T as the work habit dimension WCa corresponding to each worker, where T is the preset days, and the specific value range is greater than 1 day and less than 90 days, and T is a positive integer; The specific way to obtain the job stability dimension is: From the personnel files, the number of job changes of each worker during his / her employment time is obtained, and the inverse of the ratio between the number of job changes of each worker and his / her corresponding employment time is taken as the job stability dimension WDa of each worker; The specific method of constructing the user portrait of each worker is: Normalize the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension corresponding to each worker so that the value of each dimension is between 0 and 1. Use the normalized value of each dimension as the feature vector of the worker's user portrait, that is, the user portrait of each worker can be represented as a four-dimensional vector (WAa, WBa, WCa, WDa); form a matrix with each row representing a user portrait of a worker, and the columns corresponding to the work efficiency dimension, work skill dimension, work habit dimension, and work stability dimension, thereby constructing a user portrait of each worker; The specific method of generating the connection diagram corresponding to each employee is as follows: S1: Treat each worker as a different node, and select a worker as the target worker as the calibration node P. Sort the remaining nodes according to the worker number a in ascending order, and mark them as Ub in sequence. b refers to other different remaining nodes, and b is used as the node number of other different remaining nodes, and b is a positive integer; S2: According to the four-dimensional vector of each worker user portrait, the four-dimensional vector of each worker user portrait is used as the node vector of each node, and then the node vector of the calibration node is marked as P (DA1, DB1, DC1, DD1), and the node vectors of other different remaining nodes are marked as Ub (EAb, EBb, ECb, EDb). The similarity value Hb between the node vectors of the calibration node and other different remaining nodes is calculated by the similarity value calculation formula. S3: Nodes with similarity values ​​Hb greater than 0.9 are connected to the calibration node, and then a construction connection diagram corresponding to the calibration node is generated. Otherwise, no processing is performed and it is used as the construction connection diagram of the employee corresponding to the calibration node; S4: Repeat the above steps S1-S3 to obtain the construction connection diagram corresponding to each employee.

2. The method for constructing and clustering user portraits based on multi-dimensional attributes according to claim 1, characterized in that: The specific method to obtain different clusters of worker user portraits is: The number of construction links corresponding to each employee in the construction link graph is obtained, and workers with the same number of construction links are grouped into the same cluster. User portraits with different numbers of links are grouped into a separate cluster, thereby obtaining different clusters of worker user portraits and outputting the clustering results.

3. The method for constructing and clustering user portraits based on multi-dimensional attributes according to claim 1, characterized in that: The similarity value calculation formula is as follows: in,

Images