AI-based accurate matching method and system for flexible employment

Abstract

The invention discloses an AI-based accurate matching method and system for flexible employment, and relates to the technical field of employment allocation. Historical employment data of a user is collected, an employment data standard is set, the collected historical employment data is subjected to standardization processing, and then a data preprocessing method is adopted for processing; the method comprises the following steps: processing user historical employment data to obtain processed user historical employment data, classifying the processed user historical employment data by constructing a decision tree, extracting features of the classified user historical employment data in the decision tree through a convolutional neural network, and setting an employment matching decision based on the user historical employment data. The method comprises the following steps: selecting a plurality of decision trees, constructing a employment matching model in combination with the characteristics of each decision tree and a set matching decision, finally collecting user employment data in real time, matching the user employment data collected in real time based on the employment matching model and the matching decision, and counting the employment matching accuracy, thereby significantly improving the scientificity and practicability of user employment matching.

Description

Technical Field

[0001] This invention relates to the field of labor allocation technology, specifically to an AI-based method and system for flexible and precise labor matching. Background Technology

[0002] In the flexible employment sector, the matching between companies and job seekers has long faced numerous pain points. Traditional matching methods rely on manual screening, which lacks the ability to integrate multi-dimensional characteristics such as "job type, skill requirements, and personnel qualifications," resulting in low matching efficiency and poor accuracy. A large number of potential matching opportunities are wasted, and employment data lacks unified standards, with various employment information formats being disorganized and lacking dimensions, leading to low data utilization and difficulty in supporting accurate matching.

[0003] Existing technologies, such as the invention patent application with publication number CN119066445A, disclose a method, system, and platform for intelligent matching of skilled workers in labor services. The method includes: receiving labor demand information and analyzing the labor category and demand complexity; retrieving multiple idle workers based on the labor category, performing skill matching based on multiple vocational skill information and labor demand information, and obtaining multiple skill matching degrees; analyzing the service responsiveness of the multiple idle workers to obtain multiple service responsiveness degrees; configuring weights according to demand complexity, weighting the demand complexity, multiple service responsiveness degrees, and multiple skill matching degrees to obtain multiple talent matching degrees, and then sorting and displaying the matching results. This invention can solve the technical problem in existing methods where the complexity of the task and the flexible responsiveness of skilled workers are not fully considered, resulting in the inability to accurately match skilled workers according to labor demand. Through multi-dimensional comprehensive evaluation, it can significantly improve the efficiency and quality of talent matching.

[0004] As can be seen from the above solutions, while the current method improves matching accuracy through multi-dimensional evaluation, it still has significant shortcomings. The lack of clear quantitative standards and evaluation dimensions for the complexity of employment needs, and the vague mention of "analyzing demand complexity," easily leads to a lack of unified basis for weight allocation, affecting the objectivity of the evaluation. Summary of the Invention

[0005] The purpose of this invention is to provide an AI-based method and system for precise matching of flexible employment, which solves the problems existing in the background technology.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides an AI-based method for precise matching of flexible employment, specifically including the following steps:

[0007] S1. Collect users' historical employment data, set employment data standards, and standardize the collected historical employment data according to the set employment data standards to obtain preliminary processed historical employment data.

[0008] S2, based on the pre-processed historical employment data, the data preprocessing method is used to process the data to obtain the processed user historical employment data;

[0009] S3, based on the processed user historical employment data, classify the processed user historical employment data by constructing a decision tree, and extract the features of the classified user historical employment data in the decision tree by using a convolutional neural network to obtain the features of each decision tree;

[0010] S4, set employment matching decisions based on users' historical employment data, and construct an employment matching model by combining the characteristics of each decision tree and the set matching decisions;

[0011] S5. Collect user employment data in real time, match the collected user employment data based on the employment matching model and matching decision, and calculate the employment matching accuracy.

[0012] Preferably, the process of collecting users' historical employment data, setting employment data standards, and standardizing the collected historical employment data according to the set employment data standards to obtain preliminarily processed historical employment data includes the following steps:

[0013] Collect users' historical employment data and integrate the collected historical employment data to form a historical employment dataset;

[0014] The historical employment data includes: job type, skill requirements, and personnel qualification matching degree;

[0015] Based on business needs and industry standards, employment data standards are set. Historical employment data is then filtered according to these standards, removing data that does not meet the standards or contains missing information, resulting in pre-processed historical employment data.

[0016] Preferably, the process of processing the historical employment data based on the initially processed data using a data preprocessing method to obtain the processed user historical employment data includes the following steps:

[0017] S21, perform cleaning operations on the pre-processed historical employment data, remove duplicate records from the historical employment data, identify and remove abnormal employment data that does not conform to business logic, and obtain cleaned historical employment data.

[0018] S22, based on the cleaned historical employment data, the data on job type and skill requirement attributes in the historical employment data are encoded into a numerical form that can be directly recognized and calculated by the machine learning model;

[0019] S23, use a standardized formula to process the personnel qualification matching data of historical employment data to obtain the processed user historical employment data;

[0020] The standardized formula is as follows:

[0021] ;

[0022] in For standardized personnel qualification matching data, This data represents the matching degree of personnel qualifications to historical employment data. This represents the average of the qualification matching data for all personnel. The standard deviation of the qualification matching data for all personnel.

[0023] Preferably, the process of classifying the processed historical employment data of users by constructing decision trees and extracting the features of the classified historical employment data of users from the decision trees using a convolutional neural network to obtain the features of each decision tree includes the following steps:

[0024] S31, the processed user historical employment data is divided into two parts, one part is used as training set data and the other part is used as test set data;

[0025] S32, Build a decision tree model based on the training set data, and use the built decision tree model to perform preliminary classification of the processed user historical employment data to obtain the preliminary classified user historical employment data;

[0026] S33. The classification accuracy of the decision tree model is verified based on the test set data. The decision tree model is optimized according to the verification results to obtain the optimized decision tree model. The optimized decision tree model is then used to classify the processed user historical employment data to obtain the classified user historical employment data.

[0027] S34. Based on the optimized decision tree model, a convolutional neural network is used to extract features from the classified user historical employment data in the decision tree to obtain the features of each decision tree.

[0028] Preferably, the step of constructing a decision tree model based on the training set data and performing preliminary classification of the processed user historical employment data using the constructed decision tree model to obtain the preliminary classified user historical employment data includes the following steps:

[0029] A training sample set is established based on the training set data. The root node is established using the data in the training sample set as the sample. The data in the training sample set is then transferred to the decision tree model to complete the initialization of the decision tree model.

[0030] Based on the initialized decision tree model, starting from the root node, the optimal split is calculated using information gain as the selection criterion through the information entropy formula and the information gain formula, and the corresponding child nodes are generated.

[0031] The formula for information entropy is as follows:

[0032] ;

[0033] The information gain formula is shown below:

[0034] ;

[0035] in The input training sample set Information entropy For the training sample set belonging to the first The probability of a data sample belonging to a certain class. The number of data categories in the training sample set. Features in the training sample set Information gain Features The number of possible values, For a subset of the training sample set whose feature A takes the value v, For subset Entropy;

[0036] Repeat the above splitting operation until the output of the child nodes remains unchanged, and you will get the trained decision tree model.

[0037] The classification criteria are set as the job type and skill requirements. Based on the set classification criteria, the data in the training sample set is initially classified by the trained decision tree model to obtain the preliminary classification of user historical employment data.

[0038] Preferably, the process of validating the classification accuracy of the decision tree model based on test set data, optimizing the decision tree model according to the validation results to obtain an optimized decision tree model, and then classifying the processed user historical employment data using the optimized decision tree model to obtain the classified user historical employment data includes the following steps:

[0039] The test set data is input into the trained decision tree model, and the trained decision tree model is used for classification to obtain the classified test set data.

[0040] Set a classification accuracy threshold, and calculate the classification accuracy of the test set data based on the classification data of the test set data and the actual classification data of the test set data;

[0041] The classification accuracy of the calculated test set data is compared with the classification accuracy threshold to determine whether the set classification accuracy threshold has been reached. If the classification accuracy is lower than the set classification accuracy threshold, the decision tree model is optimized and adjusted, and the processed user historical employment data is classified through the optimized decision tree model to obtain the classified user historical employment data.

[0042] Preferably, the step of using a convolutional neural network to extract features from the processed historical employment data of users in the decision tree based on the optimized decision tree model to obtain the features corresponding to each decision tree includes the following steps:

[0043] A convolutional neural network containing convolutional layers, pooling layers, and fully connected layers is built based on the classified user historical employment data. The height and width parameters of the convolutional kernels and the weight parameters of the convolutional kernels at different positions are set in the convolutional neural network.

[0044] The categorized user historical employment data is input into a convolutional neural network, and the first convolutional layer performs convolution operations on the user historical employment data to obtain the user historical employment data features.

[0045] The formula for convolution is as follows:

[0046] ;

[0047] in For the position after convolution operation The output value at that location, For the input user's historical employment data, For the convolution kernel in The weighting parameters of the location, These are the height and width of the convolution kernel, respectively. For bias terms;

[0048] The user historical employment data features obtained after convolution are integrated and input into the pooling layer. The pooling layer performs dimensionality reduction on the received user historical employment data features, and then sends the dimensionality-reduced user historical employment data features to the fully connected layer. The fully connected layer performs feature fusion and outputs the user historical employment data features.

[0049] Preferably, the step of setting employment matching decisions based on users' historical employment data and constructing an employment matching model by combining the characteristics of each decision tree and the set matching decisions includes the following steps:

[0050] S41. Based on the obtained user historical employment data features, construct a three-dimensional feature set containing features of employment job type, skill requirements and personnel qualification matching degree, and assign initial weight coefficients to each feature in the three-dimensional feature set through the analytic hierarchy process.

[0051] S42, set matching decision rules. Based on the initial weight coefficients of each feature in the three-dimensional feature set, calculate the matching value between the user and the employment demand through Euclidean distance. The smaller the matching value, the higher the matching degree. Set a matching value threshold range, divide the priority according to the threshold in the matching, and sort them according to the priority.

[0052] The formula for calculating Euclidean distance is as follows:

[0053] ;

[0054] in The European-style distance between users and labor demand. The third feature set for user and labor needs One characteristic, The number of features in the three-dimensional feature set;

[0055] The formula for deciding on employment benefits is shown below:

[0056] ;

[0057] in, This represents the change in employment benefit decisions. Indicates user The estimated revenue parameters, Indicates user Labor costs;

[0058] Furthermore, the matching value between users and labor demand is calculated by combining labor income decision-making and Euclidean distance;

[0059] The formula for calculating the matching value is as follows:

[0060] ;

[0061] in, These are the weighting parameters for Euclidean distance and employment benefit decisions, respectively. Indicates user The matching value between the labor demand and the employment needs;

[0062] S43. Based on the set matching decision rules and the features output by each decision tree, a two-layer employment matching model is constructed: the first layer uses the matching value and priority calculated by Euclidean distance as the basis to filter out the candidate user employment list and sort it; the second layer introduces the features of the decision tree as constraints to perform secondary verification and re-sort the candidate user employment list, and outputs the final employment matching result according to the sorting.

[0063] Preferably, the real-time collection of user employment data, matching of the real-time collected user employment data based on the employment matching model and matching decisions, and statistical analysis of the employment matching accuracy include the following steps:

[0064] Real-time user employment data is collected, and the collected real-time user employment data is processed using the processing methods in S1 and S2 to obtain processed real-time user employment data.

[0065] The processed real-time user employment data is input into the employment matching model, and the matching results between users and employment needs are output according to the set matching decision rules and employment matching model.

[0066] Collect actual user employment data and summarize the actual user employment data to form an actual user employment dataset. Compare the employment matching results output by the employment matching model with the actual user employment data, and calculate the classification accuracy according to the employment job type, skill requirements and personnel qualification matching degree. Generate a matching effect analysis report to provide data support for model iteration.

[0067] This embodiment also discloses a system for a flexible employment precision matching method based on AI, including: a data collection and standardization module, a data preprocessing module, an employment matching model construction module, and an employment matching and evaluation module;

[0068] The data acquisition and standardization module is used to collect users' historical employment data and real-time employment data, integrate them to form a dataset, filter and clean the data to obtain processed user employment data.

[0069] The data preprocessing module is used to standardize the processed user employment data, encoding attribute data such as job type and skill requirements into numerical form; and to standardize the personnel qualification matching data.

[0070] The employment matching model construction module is used to combine decision tree model and convolutional neural network to build employment matching model, and set matching decision rules based on the extracted user historical employment data features to build a two-layer employment matching model to achieve accurate matching between users and employment needs.

[0071] The employment matching and evaluation module is used to output matching results based on real-time collected user employment data and employment matching model, collect actual user employment data to statistically classify accuracy, and generate a matching effect analysis report to provide data support for the iterative optimization of the model.

[0072] The beneficial effects of this invention are as follows:

[0073] (1) This invention collects users’ historical employment data, sets employment data standards, and standardizes the collected historical employment data according to the set employment data standards to obtain preliminary processed historical employment data. Then, the preliminary processed historical employment data is processed by data preprocessing methods to obtain processed user historical employment data. At the same time, based on the processed user historical employment data, decision trees are constructed to classify the processed user historical employment data, and the features of the classified user historical employment data in the decision trees are extracted by convolutional neural networks to obtain the features of each decision tree. Then, employment matching decisions are set based on user historical employment data, and employment matching models are constructed by combining the features of each decision tree and the set matching decisions. Finally, user employment data is collected in real time, and the real-time collected user employment data is matched based on the employment matching model and matching decisions. The employment matching accuracy is statistically analyzed, which significantly improves the scientificity and practicality of user employment matching.

[0074] (2) This invention transforms historical employment data into structured data with classification attributes by standardizing the processing of user employment data and extracting features, presenting the correlation between job positions, skills and qualifications, providing a basis for employment forecasting and skills reserve planning, and reducing employment costs and risks.

[0075] (3) The two-layer architecture employment matching model and matching decision mechanism constructed in this invention achieves a dual improvement in matching accuracy and scenario adaptability. Based on the three-dimensional feature set and the hierarchical analysis method, the matching decision rules can dynamically adjust the weights of job type, skill requirements and qualification matching degree according to different employment scenarios, so that the model can accurately adapt to complex and ever-changing flexible employment needs, and significantly improve the scenario coverage and decision rationality of the matching strategy. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the AI-based flexible employment precision matching method and system flow of the present invention. Detailed Implementation

[0077] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] Example 1

[0079] Please see Figure 1 This embodiment discloses an AI-based method for precise matching of flexible employment, specifically including the following steps:

[0080] S1. Collect users' historical employment data, set employment data standards, and standardize the collected historical employment data according to the set employment data standards to obtain preliminary processed historical employment data.

[0081] S2, based on the pre-processed historical employment data, the data preprocessing method is used to process the data to obtain the processed user historical employment data;

[0082] S3, based on the processed user historical employment data, classify the processed user historical employment data by constructing a decision tree, and extract the features of the classified user historical employment data in the decision tree by using a convolutional neural network to obtain the features of each decision tree;

[0083] S4, set employment matching decisions based on users' historical employment data, and construct an employment matching model by combining the characteristics of each decision tree and the set matching decisions;

[0084] S5. Collect user employment data in real time, match the collected user employment data based on the employment matching model and matching decision, and calculate the employment matching accuracy.

[0085] Further, please refer to Figure 1 The process involves collecting users' historical employment data, setting employment data standards, and standardizing the collected historical employment data according to these standards to obtain pre-processed historical employment data. This includes the following steps:

[0086] Collect users' historical employment data and integrate the collected historical employment data to form a historical employment dataset;

[0087] The historical employment data includes: job type, skill requirements, and personnel qualification matching degree;

[0088] Based on business needs and industry standards, employment data standards are set. Historical employment data is then filtered according to these standards, removing data that does not meet the standards or contains missing information, resulting in pre-processed historical employment data.

[0089] Further, please refer to Figure 1 Based on the initially processed historical employment data, data preprocessing methods are used to process the data to obtain the processed user historical employment data, including the following steps:

[0090] S21, perform cleaning operations on the pre-processed historical employment data, remove duplicate records from the historical employment data, identify and remove abnormal employment data that does not conform to business logic, and obtain cleaned historical employment data.

[0091] S22, based on the cleaned historical employment data, the data on job type and skill requirement attributes in the historical employment data are encoded into a numerical form that can be directly recognized and calculated by the machine learning model;

[0092] S23, use a standardized formula to process the personnel qualification matching data of historical employment data to obtain the processed user historical employment data;

[0093] The standardized formula is as follows:

[0094] ;

[0095] in For standardized personnel qualification matching data, This data represents the matching degree of personnel qualifications to historical employment data. This represents the average of the qualification matching data for all personnel. The standard deviation of the qualification matching data for all personnel.

[0096] Further, please refer to Figure 1 Based on the processed historical employment data of users, decision trees are constructed to classify the processed historical employment data of users, and features of the classified historical employment data of users in the decision trees are extracted by convolutional neural networks. The features of each decision tree are obtained by the following steps:

[0097] S31, the processed user historical employment data is divided into two parts, one part is used as training set data and the other part is used as test set data;

[0098] S32, Build a decision tree model based on the training set data, and use the built decision tree model to perform preliminary classification of the processed user historical employment data to obtain the preliminary classified user historical employment data;

[0099] S33. The classification accuracy of the decision tree model is verified based on the test set data. The decision tree model is optimized according to the verification results to obtain the optimized decision tree model. The optimized decision tree model is then used to classify the processed user historical employment data to obtain the classified user historical employment data.

[0100] S34. Based on the optimized decision tree model, a convolutional neural network is used to extract features from the classified user historical employment data in the decision tree to obtain the features of each decision tree.

[0101] Further, please refer to Figure 1 A decision tree model is constructed based on the training set data, and the processed user historical employment data is initially classified using the constructed decision tree model. The initial classification of user historical employment data includes the following steps:

[0102] A training sample set is established based on the training set data. The root node is established using the data in the training sample set as the sample. The data in the training sample set is then transferred to the decision tree model to complete the initialization of the decision tree model.

[0103] Based on the initialized decision tree model, starting from the root node, the optimal split is calculated using information gain as the selection criterion through the information entropy formula and the information gain formula, and the corresponding child nodes are generated.

[0104] The formula for information entropy is as follows:

[0105] ;

[0106] The information gain formula is shown below:

[0107] ;

[0108] in The input training sample set Information entropy For the training sample set belonging to the first The probability of a data sample belonging to a certain class. The number of data categories in the training sample set. Features in the training sample set Information gain Features The number of possible values, For a subset of the training sample set whose feature A takes the value v, For subset Entropy;

[0109] Repeat the above splitting operation until the output of the child nodes remains unchanged, and you will get the trained decision tree model.

[0110] The classification criteria are set as the job type and skill requirements. Based on the set classification criteria, the data in the training sample set is initially classified by the trained decision tree model to obtain the preliminary classification of user historical employment data.

[0111] Further, please refer to Figure 1 The classification accuracy of the decision tree model is verified based on the test set data. The model is then optimized based on the verification results to obtain an optimized decision tree model. This optimized model is then used to classify the processed historical employment data of users, resulting in the classified historical employment data of users. The steps include:

[0112] The test set data is input into the trained decision tree model, and the trained decision tree model is used for classification to obtain the classified test set data.

[0113] Set a classification accuracy threshold, and calculate the classification accuracy of the test set data based on the classification data of the test set data and the actual classification data of the test set data;

[0114] The classification accuracy of the calculated test set data is compared with the classification accuracy threshold to determine whether the set classification accuracy threshold has been reached. If the classification accuracy is lower than the set classification accuracy threshold, the decision tree model is optimized and adjusted, and the processed user historical employment data is classified through the optimized decision tree model to obtain the classified user historical employment data.

[0115] Further, please refer to Figure 1 Based on the optimized decision tree model, a convolutional neural network is used to extract features from the processed historical employment data of users in the decision tree to obtain the features corresponding to each decision tree, including the following steps:

[0116] A convolutional neural network containing convolutional layers, pooling layers, and fully connected layers is built based on the classified user historical employment data. The height and width parameters of the convolutional kernels and the weight parameters of the convolutional kernels at different positions are set in the convolutional neural network.

[0117] The categorized user historical employment data is input into a convolutional neural network, and the first convolutional layer performs convolution operations on the user historical employment data to obtain the user historical employment data features.

[0118] The formula for convolution is as follows:

[0119] ;

[0120] in For the position after convolution operation The output value at that location, For the input user's historical employment data, For the convolution kernel in The weighting parameters of the location, These are the height and width of the convolution kernel, respectively. For bias terms;

[0121] The user historical employment data features obtained after convolution are integrated and input into the pooling layer. The pooling layer performs dimensionality reduction on the received user historical employment data features, and then sends the dimensionality-reduced user historical employment data features to the fully connected layer. The fully connected layer performs feature fusion and outputs the user historical employment data features.

[0122] Further, please refer to Figure 1 The process of setting employment matching decisions based on users' historical employment data and constructing an employment matching model by combining the characteristics of each decision tree and the set matching decisions includes the following steps:

[0123] S41. Based on the obtained user historical employment data features, construct a three-dimensional feature set containing features of employment job type, skill requirements and personnel qualification matching degree, and assign initial weight coefficients to each feature in the three-dimensional feature set through the analytic hierarchy process.

[0124] S42, set matching decision rules. Based on the initial weight coefficients of each feature in the three-dimensional feature set, calculate the matching value between the user and the employment demand through Euclidean distance. The smaller the matching value, the higher the matching degree. Set a matching value threshold range, divide the priority according to the threshold in the matching, and sort them according to the priority.

[0125] The formula for calculating Euclidean distance is as follows:

[0126] ;

[0127] in The European-style distance between users and labor demand. The third feature set for user and labor needs One characteristic, The number of features in the three-dimensional feature set;

[0128] The formula for deciding on employment benefits is shown below:

[0129] ;

[0130] in, This represents the change in employment benefit decisions. Indicates user The estimated revenue parameters, Indicates user Labor costs;

[0131] Furthermore, the matching value between users and labor demand is calculated by combining labor income decision-making and Euclidean distance;

[0132] The formula for calculating the matching value is as follows:

[0133] ;

[0134] in, These are the weighting parameters for Euclidean distance and employment benefit decisions, respectively. Indicates user The matching value between the labor demand and the employment needs;

[0135] S43. Based on the set matching decision rules and the features output by each decision tree, a two-layer employment matching model is constructed: the first layer uses the matching value and priority calculated by Euclidean distance as the basis to filter out the candidate user employment list and sort it; the second layer introduces the features of the decision tree as constraints to perform secondary verification and re-sort the candidate user employment list, and outputs the final employment matching result according to the sorting.

[0136] Further, please refer to Figure 1 The process involves collecting user employment data in real time, matching the collected data based on an employment matching model and matching decisions, and calculating the employment matching accuracy. This includes the following steps:

[0137] Real-time user employment data is collected, and the collected real-time user employment data is processed using the processing methods in S1 and S2 to obtain processed real-time user employment data.

[0138] The processed real-time user employment data is input into the employment matching model, and the matching results between users and employment needs are output according to the set matching decision rules and employment matching model.

[0139] Collect actual user employment data and summarize the actual user employment data to form an actual user employment dataset. Compare the employment matching results output by the employment matching model with the actual user employment data, and calculate the classification accuracy according to the employment job type, skill requirements and personnel qualification matching degree. Generate a matching effect analysis report to provide data support for model iteration.

[0140] Example 2

[0141] This embodiment also discloses a system for a flexible employment precision matching method based on AI, including: a data collection and standardization module, a data preprocessing module, an employment matching model construction module, and an employment matching and evaluation module;

[0142] The data acquisition and standardization module is used to collect users' historical employment data and real-time employment data, integrate them to form a dataset, filter and clean the data to obtain processed user employment data.

[0143] The data preprocessing module is used to standardize the processed user employment data, encoding attribute data such as job type and skill requirements into numerical form; and to standardize the personnel qualification matching data.

[0144] The employment matching model construction module is used to combine decision tree model and convolutional neural network to build employment matching model, and set matching decision rules based on the extracted user historical employment data features to build a two-layer employment matching model to achieve accurate matching between users and employment needs.

[0145] The employment matching and evaluation module is used to output matching results based on real-time collected user employment data and employment matching model, collect actual user employment data to statistically classify accuracy, and generate a matching effect analysis report to provide data support for the iterative optimization of the model.

[0146] It should be noted that

[0147] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A flexible employment precision matching method based on AI, characterized in that, Includes the following steps: S1. Collect users' historical employment data, set employment data standards, and standardize the collected historical employment data according to the set employment data standards to obtain preliminary processed historical employment data. S2, based on the pre-processed historical employment data, the data preprocessing method is used to process the data to obtain the processed user historical employment data; S3, based on the processed user historical employment data, classify the processed user historical employment data by constructing a decision tree, and extract the features of the classified user historical employment data in the decision tree by using a convolutional neural network to obtain the features of each decision tree; S4, set employment matching decisions based on users' historical employment data, and construct an employment matching model by combining the characteristics of each decision tree and the set matching decisions; S5. Collect user employment data in real time, match the collected user employment data based on the employment matching model and matching decision, and calculate the employment matching accuracy.

2. The AI-based flexible employment precision matching method according to claim 1, characterized in that, The process of collecting users' historical employment data, setting employment data standards, and standardizing the collected historical employment data according to the set employment data standards to obtain preliminarily processed historical employment data includes the following steps: Collect users' historical employment data and integrate the collected historical employment data to form a historical employment dataset; The historical employment data includes: job type, skill requirements, and personnel qualification matching degree; Based on business needs and industry standards, employment data standards are set. Historical employment data is then filtered according to these standards, removing data that does not meet the standards or contains missing information, resulting in pre-processed historical employment data.

3. The AI-based flexible employment precision matching method according to claim 1, characterized in that, The process of preprocessing the historical employment data, based on the initially processed data, to obtain the processed user historical employment data includes the following steps: S21, perform a cleaning operation on the pre-processed historical employment data, remove duplicate records from the historical employment data, identify and remove abnormal employment data that does not conform to business logic, and obtain the cleaned historical employment data. S22, based on the cleaned historical employment data, the data on job type and skill requirement attributes in the historical employment data are encoded into a numerical form that can be directly recognized and calculated by the machine learning model; S23, use a standardized formula to process the personnel qualification matching data of historical employment data to obtain the processed user historical employment data; The standardized formula is as follows: ； in For standardized personnel qualification matching data, This data represents the matching degree of personnel qualifications based on historical employment data. This represents the average of the qualification matching data for all personnel. The standard deviation of the qualification matching data for all personnel.

4. The AI-based flexible employment precision matching method according to claim 1, characterized in that, The process involves classifying the processed historical employment data of users by constructing decision trees, and extracting features of the classified historical employment data from the decision trees using a convolutional neural network. The process for obtaining the features of each decision tree includes the following steps: S31, the processed user historical employment data is divided into two parts, one part is used as training set data and the other part is used as test set data; S32, Build a decision tree model based on the training set data, and use the built decision tree model to perform preliminary classification of the processed user historical employment data to obtain the preliminary classified user historical employment data; S33. The classification accuracy of the decision tree model is verified based on the test set data. The decision tree model is optimized according to the verification results to obtain the optimized decision tree model. The optimized decision tree model is then used to classify the processed user historical employment data to obtain the classified user historical employment data. S34. Based on the optimized decision tree model, a convolutional neural network is used to extract features from the classified user historical employment data in the decision tree to obtain the features of each decision tree.

5. The AI-based flexible employment precision matching method according to claim 4, characterized in that, The process of constructing a decision tree model based on the training set data and then using the constructed decision tree model to perform preliminary classification of the processed user historical employment data to obtain the preliminary classified user historical employment data includes the following steps: A training sample set is established based on the training set data. The root node is established using the data in the training sample set as the sample. The data in the training sample set is then transferred to the decision tree model to complete the initialization of the decision tree model. Based on the initialized decision tree model, starting from the root node, the optimal split is calculated using information gain as the selection criterion through the information entropy formula and the information gain formula, and the corresponding child nodes are generated. The formula for information entropy is as follows: ； The information gain formula is shown below: ； in The input training sample set Information entropy For the training sample set belonging to the first The probability of a data sample belonging to a certain class. The number of data categories in the training sample set. Features in the training sample set Information gain Features The number of possible values, For a subset of the training sample set whose feature A takes the value v, For subset Entropy; Repeat the above splitting operation until the output of the child nodes remains unchanged, and you will get the trained decision tree model. The classification criteria are set as the job type and skill requirements. Based on the set classification criteria, the data in the training sample set is initially classified by the trained decision tree model to obtain the preliminary classification of user historical employment data.

6. The AI-based flexible employment precision matching method according to claim 4, characterized in that, The process of validating the classification accuracy of the decision tree model based on test set data, optimizing the decision tree model based on the validation results, obtaining an optimized decision tree model, and then classifying the processed user historical employment data using the optimized decision tree model to obtain the classified user historical employment data includes the following steps: The test set data is input into the trained decision tree model, and the trained decision tree model is used for classification to obtain the classified test set data. Set a classification accuracy threshold, and calculate the classification accuracy of the test set data based on the classification data of the test set data and the actual classification data of the test set data; The classification accuracy of the calculated test set data is compared with the classification accuracy threshold to determine whether the set classification accuracy threshold has been reached. If the classification accuracy is lower than the set classification accuracy threshold, the decision tree model is optimized and adjusted, and the processed user historical employment data is classified through the optimized decision tree model to obtain the classified user historical employment data.

7. The AI-based flexible employment precise matching method according to claim 1, characterized in that, The optimized decision tree model employs a convolutional neural network to extract features from the processed historical employment data of users in the decision tree, obtaining the features corresponding to each decision tree, including the following steps: A convolutional neural network containing convolutional layers, pooling layers, and fully connected layers is built based on the classified user historical employment data. The height and width parameters of the convolutional kernels and the weight parameters of the convolutional kernels at different positions are set in the convolutional neural network. The categorized user historical employment data is input into a convolutional neural network, and the first convolutional layer performs convolution operations on the user historical employment data to obtain the user historical employment data features. The formula for convolution is as follows: ； in For the position after convolution operation The output value at that location, For the input user's historical employment data, For the convolution kernel in The weighting parameters of the location, These are the height and width of the convolution kernel, respectively. For bias terms; The user historical employment data features obtained after convolution are integrated and input into the pooling layer. The pooling layer performs dimensionality reduction on the received user historical employment data features, and then sends the dimensionality-reduced user historical employment data features to the fully connected layer. The fully connected layer performs feature fusion and outputs the user historical employment data features.

8. The AI-based flexible employment precision matching method according to claim 1, characterized in that, The process of setting employment matching decisions based on users' historical employment data and constructing an employment matching model by combining the characteristics of each decision tree and the set matching decisions includes the following steps: S41. Based on the obtained user historical employment data features, construct a three-dimensional feature set containing features of employment job type, skill requirements and personnel qualification matching degree, and assign initial weight coefficients to each feature in the three-dimensional feature set through the analytic hierarchy process. S42, set matching decision rules. Based on the initial weight coefficients of each feature in the three-dimensional feature set, calculate the matching value between users and employment needs through Euclidean distance and employment benefit decision. The smaller the matching value, the higher the matching degree. Set a matching value threshold range, divide the priority according to the threshold in the matching, and sort them according to the priority. The formula for calculating Euclidean distance is as follows: ； in The European-style distance between users and labor demand. The third feature set for user and labor needs One characteristic, The number of features in the three-dimensional feature set; The formula for deciding on employment benefits is shown below: ； in, This represents the change in employment benefit decisions. Indicates user The estimated revenue parameters, Indicates user Labor costs; The matching value between users and labor demand is calculated by combining labor income decision-making and Euclidean distance. The formula for calculating the matching value is as follows: ； in, These are the weighting parameters for Euclidean distance and employment benefit decisions, respectively. Indicates user The matching value between the labor demand and the employment needs; S43. Based on the set matching decision rules and the features output by each decision tree, a two-layer employment matching model is constructed: the first layer uses the matching value and priority calculated by Euclidean distance as the basis to filter out the candidate user employment list and sort it; the second layer introduces the features of the decision tree as constraints to perform secondary verification and re-sort the candidate user employment list, and outputs the final employment matching result according to the sorting.

9. The AI-based flexible employment precision matching method according to claim 1, characterized in that, The process of collecting user employment data in real time, matching the collected user employment data based on the employment matching model and matching decisions, and calculating the employment matching accuracy includes the following steps: Real-time user employment data is collected, and the collected real-time user employment data is processed using the processing methods in S1 and S2 to obtain processed real-time user employment data. The processed real-time user employment data is input into the employment matching model, and the matching results between users and employment needs are output according to the set matching decision rules and employment matching model. Collect actual user employment data and summarize the actual user employment data to form an actual user employment dataset. Compare the employment matching results output by the employment matching model with the actual user employment data, and calculate the classification accuracy according to the employment job type, skill requirements and personnel qualification matching degree. Generate a matching effect analysis report to provide data support for model iteration.

10. A system for implementing the AI-based flexible employment precision matching method as described in claims 1-9, characterized in that, include: The module includes data acquisition and standardization, data preprocessing, employment matching model construction, and employment matching and evaluation. The data acquisition and standardization module is used to collect users' historical employment data and real-time employment data, integrate them to form a dataset, filter and clean the data to obtain processed user employment data. The data preprocessing module is used to standardize the processed user employment data, encoding attribute data such as job type and skill requirements into numerical form; and to standardize the personnel qualification matching data. The employment matching model construction module is used to combine decision tree model and convolutional neural network to build employment matching model, and set matching decision rules based on the extracted user historical employment data features to build a two-layer employment matching model to achieve accurate matching between users and employment needs. The employment matching and evaluation module is used to output matching results based on real-time collected user employment data and employment matching model, collect actual user employment data to statistically classify accuracy, and generate a matching effect analysis report to provide data support for the iterative optimization of the model.

Images