Information matching method, recruitment system testing method and system, terminal and medium

Abstract

The invention provides an information matching method, a recruitment system testing method, a recruitment system testing system, a terminal and a medium. The method comprises the following steps: respectively encoding an extracted job seeker feature vector and an extracted recruiter feature vector into a job seeker behavior code and a recruiter behavior code; based on a preset gradient descent algorithm, importance weights of the job seeker behavior codes and the recruiter behavior codes in final matching are determined in an adaptive learning mode; and fusing the weighted job seeker behavior code and the weighted recruiter behavior code, and calculating a matching degree between the comprehensive matching code of the job seeker obtained by fusion and the feature code of the potential matching object so as to determine an optimal matching result between the recruiter and the job seeker based on the matching degree. According to the invention, through a dynamic adaptive learning weight distribution mechanism and a behavior coding fusion strategy, independent preferences of the job seeker and the recruiter can be fully considered and satisfied, and the accuracy of the matching process can be ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of information processing and software testing, and particularly relates to an information matching method, a recruitment system testing method, a system, a terminal and a medium. Background Art

[0002] An online recruitment system is an Internet-based recruitment platform that allows enterprises to post job vacancy information and allows job seekers to search for and apply for these jobs via the Internet.

[0003] However, most of the existing online recruitment systems currently adopt fixed matching algorithms and cannot adjust the matching strategy according to the dynamic characteristics of job seekers and recruiters, resulting in poor matching effects. Moreover, the feature weights of job seekers and recruiters in traditional online recruitment systems are usually preset and fixed and cannot be optimized according to the actual matching effect, affecting the accuracy and personalization of the matching.

[0004] Therefore, there are defects in the prior art and it needs to be improved and developed. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an information matching method, a recruitment system testing method, a system, a terminal and a medium, which can fully consider and meet the independent preferences of both job seekers and recruiters and ensure the accuracy of the matching process.

[0006] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0007] An information matching method, wherein the method includes:

[0008] Obtain the independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector;

[0009] Encode the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively;

[0010] Based on a preset gradient descent algorithm, determine the importance weights of the job seeker behavior code and the recruiter behavior code in the final matching through an adaptive learning method to obtain a weighted job seeker behavior code and a weighted recruiter behavior code;

[0011] Fuse the weighted job seeker behavior code and the weighted recruiter behavior code to obtain the comprehensive matching code of the job seeker, and calculate the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object to determine the best matching result between the recruiter and the job seeker based on the matching degree.

[0012] In one implementation, obtaining the independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector includes:

[0013] Extracting the independent feature data of the job seeker from the detailed features of the job seeker to obtain a job seeker feature vector; the detailed features of the job seeker include the activity of the job seeker on the recruitment platform, the career development of the job seeker, personal information, and job hunting preferences;

[0014] Extracting the independent feature data of the recruiter from the detailed features of the recruiter to obtain a recruiter feature vector; the detailed features of the recruiter include basic enterprise information, corporate culture and values, historical recruitment records, job requirements, corporate reputation and brand image, financial status, and development prospects.

[0015] In one implementation, encoding the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively includes:

[0016] Encoding the job seeker feature vector into a job seeker behavior code by using a job seeker encoding model constructed based on a first deep neural network model;

[0017] Encoding the recruiter feature vector into a recruiter behavior code by using a recruiter encoding model constructed based on a second deep neural network model.

[0018] In one implementation, fusing the weighted job seeker behavior code and the weighted recruiter behavior code to obtain the comprehensive matching code of the job seeker includes:

[0019] Fusing the weighted job seeker behavior code and the weighted recruiter behavior code by using a preset weighted summation method to obtain the comprehensive matching code of the job seeker;

[0020] wherein, the preset weighted summation method is: C comb = ω j · C j + ω e · C e ;

[0021] wherein, C comb represents the comprehensive matching code, C j represents the job seeker behavior code, C e represents the recruiter behavior code, ω j represents the importance weight corresponding to the job seeker behavior code, ω e represents the importance weight corresponding to the recruiter behavior code.

[0022] In one implementation, calculating the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object includes:

[0023] Calculating the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object by using the cosine similarity calculation method.

[0024] In one implementation, the information matching method further includes:

[0025] Extracting the English application ability feature vector from the English test results of the job seeker;

[0026] Determining the importance weight of the English application ability feature vector in the final matching based on the reinforcement learning algorithm and the gradient descent algorithm to obtain the weighted English application ability feature vector;

[0027] Among them, the step of fusing the weighted job seeker behavior code and the weighted recruiter behavior code to obtain the comprehensive matching code of the job seeker includes:

[0028] Fusing the weighted job seeker behavior code, the weighted English application ability feature vector and the weighted recruiter behavior code to obtain the comprehensive matching code of the job seeker.

[0029] The present invention also discloses a recruitment system testing method, wherein the method includes:

[0030] Slicing the multi-platform recruitment system based on the feature dimension to obtain the recruitment systems corresponding to different slice dimensions of each platform;

[0031] Testing and analyzing the matching effects corresponding to the recruitment systems of each slice dimension based on the above-mentioned information matching method to obtain the test results corresponding to the multi-platform recruitment system.

[0032] The present invention also discloses an information matching system, wherein the system includes:

[0033] A feature determination module, configured to obtain the independent feature data of the job seeker and the recruiter to obtain the job seeker feature vector and the recruiter feature vector;

[0034] A feature encoding module, configured to encode the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively;

[0035] A weight determination module, configured to determine the importance weights of the job seeker behavior code and the recruiter behavior code in the final matching based on a preset gradient descent algorithm and in an adaptive learning manner to obtain the weighted job seeker behavior code and the weighted recruiter behavior code;

[0036] A fusion module, configured to fuse the weighted job seeker behavior encoding and the weighted recruiter behavior encoding to obtain the comprehensive matching encoding of the job seeker;

[0037] A matching module, configured to calculate the matching degree between the comprehensive matching encoding of the job seeker and the feature encoding of a potential matching object, and determine the best matching result between the recruiter and the job seeker based on the matching degree.

[0038] The present invention also discloses a terminal, which includes: a memory, a processor, and an information matching program stored on the memory and executable on the processor. When the information matching program is executed by the processor, the steps of the information matching method described above are implemented.

[0039] The present invention also discloses a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and the computer program can be executed to implement the steps of the information matching method described above.

[0040] The information matching method, recruitment system testing method, system, terminal and medium provided by the present invention. The information matching method includes: obtaining independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector; encoding the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively; based on a preset gradient descent algorithm, and by means of adaptive learning, determining the importance weights of the job seeker behavior code and the recruiter behavior code in the final matching to obtain a weighted job seeker behavior code and a weighted recruiter behavior code; fusing the weighted job seeker behavior code and the weighted recruiter behavior code to obtain a comprehensive matching code of the job seeker, and calculating the matching degree between the comprehensive matching code of the job seeker and the feature code of a potential matching object to determine the best matching result between the recruiter and the job seeker based on the matching degree. It can be seen that the present invention encodes the job seeker feature vector and the recruiter feature vector corresponding to job seekers and recruiters into a job seeker behavior code and a recruiter behavior code respectively, and then allocates the feature weights of job seekers and recruiters through a dynamic adaptive learning weight allocation mechanism, so that the weights can be dynamically adjusted to optimize the matching effect. Furthermore, through a behavior code fusion strategy, the weighted job seeker behavior code and the weighted recruiter behavior code are fused, and then the matching degree between the comprehensive matching code of the job seeker obtained by fusion and the feature code of a potential matching object is calculated to determine the best matching result between the recruiter and the job seeker based on the matching degree. That is, the information matching implemented by the present application through a dynamic adaptive learning weight allocation mechanism and a behavior code fusion strategy can fully consider and meet the independent preferences of both job seekers and recruiters, and can ensure the accuracy of the matching process between recruiters and job seekers. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of a preferred embodiment of the information matching method in the present invention;

[0042] Figure 2 is a logical schematic block diagram of a preferred embodiment of the recruitment system testing method in the present invention;

[0043] Figure 3 is a schematic diagram of a recruitment system testing framework disclosed by the present invention;

[0044] Figure 4 is a functional principle block diagram of a preferred embodiment of the information matching system in the present invention;

[0045] Figure 5 is a functional principle block diagram of a preferred embodiment of the terminal in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To make the objectives, technical solutions and advantages of the present invention more clear and definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific examples described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] Please refer to Figure 1 , Figure 1 which is a flowchart of the information matching method in the present invention. As Figure 1 shown, the information matching method described in the embodiments of the present invention includes:

[0048] Step S11: Obtain the independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector.

[0049] In this embodiment, first, the features of job seekers and recruiters are extracted, that is, the independent feature data of job seekers and recruiters are obtained to obtain a job seeker feature vector and a recruiter feature vector.

[0050] Specifically, for the feature extraction of job seekers, the independent feature data of job seekers are extracted from the detailed features of job seekers to obtain a job seeker feature vector; the detailed features of job seekers include the activity of job seekers on the recruitment platform, the career development of job seekers, personal information, and job hunting preferences.

[0051] For example, features are extracted from the detailed features of job seekers to form a job seeker feature vector F j , and the specific details of the detailed features of job seekers include: the activity of job seekers on the recruitment platform, such as the frequency of job seekers logging in to the recruitment platform, the browsing records of job seekers on the platform, the resume submission records of job seekers, and the interaction behaviors of job seekers on the platform; the career development of job seekers, such as the education background, major, school information, work experience, skill certificates obtained by job seekers, and project experience participated by job seekers; the personal information of job seekers, such as the age range, gender, city and region where job seekers are located, industry or field, and expected salary range; the job hunting preferences of job seekers, such as the preferred job type, work nature, work location, working hours, and company scale.

[0052] Specifically, for the feature extraction of recruiters, the independent feature data of recruiters are extracted from the detailed features of recruiters to obtain a recruiter feature vector; the detailed features of recruiters include basic enterprise information, corporate culture and values, historical recruitment records, job requirements, corporate reputation and brand image, financial status, and development prospects.

[0053] For example, features are extracted from the detailed features of recruiters to form a recruiter feature vector F e, the detailed features specifically include: basic enterprise information, such as the number of employees, the number of branch companies, the industry or professional field to which the enterprise belongs, the length of time since the enterprise was established, the main office location of the enterprise, the business coverage area, and the enterprise type; corporate culture and values, such as the long-term goals and values of the enterprise, the working environment, team atmosphere, and management style of the enterprise; historical recruitment records, such as the frequency of job postings by the enterprise, the types of positions the enterprise often recruits for, the proportion of successful recruitment after the enterprise posts a position, and the average length of time employees work in the enterprise; job requirements, such as the requirements for job seekers' skills, experience, and educational background, the salary level and welfare benefits provided, and the geographical location of the position; enterprise reputation and brand image, such as the reputation and brand influence of the enterprise in the external market, the evaluations and feedback from existing employees on the enterprise, and the evaluations of the enterprise's products or services by customers; financial status, such as the annual operating income of the enterprise, financial indicators such as the profit margin and growth rate of the enterprise, and the financing or investment situation obtained by the enterprise; development prospects, such as the market position and competitiveness of the enterprise in the industry, the development plan and strategic direction of the enterprise, and the R & D investment and innovation ability of the enterprise.

[0054] In this embodiment, it may also specifically include: extracting an English application ability feature vector from the English test results of job seekers. It can be understood that by integrating a third-party English test service APP on the online recruitment system, the English test is integrated into the recruitment process so that job seekers can directly take the test on the recruitment platform. After the test is completed, the corresponding English scores are automatically given, and the English application ability of job seekers can be extracted from the English test results feedback by the third-party English test service APP.

[0055] It should be noted that in addition to the existing feature data of job seekers and recruiters, more diverse dimensional feature data can be considered, such as social media behavior data, online evaluation data, professional skill certification data, etc., to further improve the accuracy of matching.

[0056] Step S12: Encode the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively.

[0057] In this embodiment, the feature vectors corresponding to job seekers and recruiters are encoded into their corresponding behavior codes respectively. Specifically, a job seeker encoding model constructed based on the first deep neural network model is used to encode the job seeker feature vector into a job seeker behavior code, and a recruiter encoding model constructed based on the second deep neural network model is used to encode the recruiter feature vector into a recruiter behavior code. It can be understood that the feature vectors of job seekers and recruiters are encoded into behavior codes using their respective independent encoding models.

[0058] For example, using a job seeker encoding model constructed based on a deep neural network (DNN), the job seeker feature vector F j is encoded into the job seeker behavior encoding C j . Among them, in the input layer of the model, the number of neurons in the input layer is defined to match the dimension of the job seeker feature vector; in the hidden layer, 128 neurons can be set on the first hidden layer, and the ReLU (Rectified Linear Unit) can be used as the activation function. Then, more hidden layers can be added, and the ReLU can also be used as the activation function after each hidden layer; in the regularization layer, a Dropout layer can be added after each hidden layer, and the Dropout ratio can be set to 0.5; in the output layer, the number of neurons in the output layer is defined. Among them, for the behavior encoding of classification problems, the Softmax activation function is used; for the behavior encoding of regression problems, the linear activation function is used. Similar to the job seeker encoding model, the recruiter encoding model is also an encoding model constructed based on a deep neural network, that is, the recruiter feature vector F e is also encoded into the recruiter behavior encoding C e . More advanced encoding technologies, such as natural language processing (NLP) technology, can also be used to perform more refined encoding on the feature data of job seekers and recruiters, improving the accuracy and interpretability of behavior encoding.

[0059] Step S13: Based on a preset gradient descent algorithm, and by means of adaptive learning, determine the importance weights of the job seeker behavior encoding and the recruiter behavior encoding in the final matching, and obtain the weighted job seeker behavior encoding and the weighted recruiter behavior encoding.

[0060] In this embodiment, the feature weights of job seekers and recruiters are assigned through a dynamic adaptive learning weight allocation mechanism, that is, based on a preset gradient descent algorithm, and by means of adaptive learning, the importance weights of the job seeker behavior encoding and the recruiter behavior encoding in the final matching are assigned, and the weighted job seeker behavior encoding and the weighted recruiter behavior encoding are obtained. It can be understood that through the gradient descent method, according to historical matching data, the importance weights ω j and ω e of the job seeker and recruiter features in the final matching are adaptively learned, so that the weights can be dynamically adjusted to optimize the matching effect.

[0061] For example, first, initialize the weights, that is, assign an initial weight to each feature in the feature vectors corresponding to job seekers and recruiters. For example, initialize the weights to small random values, and then select a loss function according to the initial weights to measure the difference between the predicted matching result and the actual matching result. Among them, for classification problems, cross-entropy loss can be used; for regression problems, mean squared error can be used.

[0062] Moreover, for each weight parameter, calculate the gradient of the loss function with respect to this parameter through the backpropagation algorithm, that is, calculate the partial derivative of the loss function with respect to each weight.

[0063] Then, use the gradient descent algorithm to update the weights. The specific steps are as follows:

[0064] Determine the learning rate η, which is a hyperparameter that controls the step size of weight update.

[0065] For each weight parameter ω, that is:

[0066]

[0067] Among them, represents the gradient of the loss function L with respect to the weight ω.

[0068] It should also be noted that for the feature vector of English application ability extracted from the English test results of job seekers, the importance weight of the feature vector of English application ability in the final matching can be determined based on the reinforcement learning algorithm and the gradient descent algorithm to obtain the weighted feature vector of English application ability. It can be understood that the method of combining the reinforcement learning algorithm and the gradient descent algorithm can be used to learn and optimize the importance weight of the feature vector of English application ability in the final matching.

[0069] For example, in the reinforcement learning algorithm, the state consists of the feature vector of the English application ability of the job seeker and other relevant features, and these features jointly describe the comprehensive ability of the job seeker. The action refers to the adjustment of the corresponding weight of the feature vector of the English application ability during the matching process, which includes increasing or decreasing the weight of the English ability to optimize the matching result. Moreover, the reward function gives feedback based on the matching result. If the matching is successful, that is, when the job seeker is selected by the recruiter and successfully employed, a positive reward is given; if the matching fails, that is, when the job seeker is not selected or performs poorly after employment, a negative reward is given. Then, the rule for selecting actions in a given state is defined by randomly initializing the policy. For each matching case, perform the following steps:

[0070] Select an action to adjust the weights corresponding to the English application ability feature vector according to the current policy, execute the selected action, that is, adjust the weights, observe the matching result, then calculate the reward according to the reward function, and then use Q-learning to update the policy. Among them, Q-learning estimates the expected return of taking action a in state s by learning the value function Q(s, a), so as to be more inclined to select actions that lead to high rewards in similar states. In the gradient descent algorithm, calculate the loss function between the weighted English application ability feature vector and the actual matching result. This loss function reflects the difference between the predicted match and the actual match, that is, use the gradient descent algorithm to optimize the English application ability weight parameters to minimize the loss function. By iteratively adjusting the weights, the optimal weight configuration of the English application ability feature vector can be determined to achieve the best matching effect. It can be understood that more advanced machine learning algorithms, such as deep learning and reinforcement learning, can be used to optimize the weight allocation strategy so that the system can adapt to market changes and user needs more quickly.

[0071] Step S14: Fuse the weighted job seeker behavior encoding and the weighted recruiter behavior encoding to obtain the comprehensive matching encoding of the job seeker, and calculate the matching degree between the comprehensive matching encoding of the job seeker and the feature encoding of the potential matching object to determine the best matching result between the recruiter and the job seeker based on the matching degree.

[0072] In this embodiment, the weighted job seeker and recruiter behavior encodings are fused into a comprehensive matching encoding. Specifically, the weighted job seeker behavior encoding and the weighted recruiter behavior encoding are fused using a preset weighted summation method to obtain the comprehensive matching encoding of the job seeker; among them, the preset weighted summation method is:

[0073] Ccomb = ωj·Cj + ωe·Ce;

[0074] where C comb represents the comprehensive matching encoding, C j represents the job seeker behavior encoding, C e represents the recruiter behavior encoding, ω j represents the importance weight corresponding to the job seeker behavior encoding, ω e represents the importance weight corresponding to the recruiter behavior encoding.

[0075] In this embodiment, the feature vectors of job seekers and recruiters are extracted. The feature vector of English application ability is extracted from the English test results of job seekers, and the corresponding feature vectors of job seekers and recruiters are respectively encoded into their corresponding behavior codes. Then, through a dynamic adaptive learning weight allocation mechanism, the feature weights of job seekers and recruiters are allocated, and after determining the optimal weight configuration of the English application ability feature vector through a reinforcement learning algorithm and a gradient descent algorithm, it may further specifically include fusing the weighted behavior codes of job seekers, the weighted feature vector of English application ability, and the weighted behavior codes of recruiters to obtain the comprehensive matching code of job seekers.

[0076] In this embodiment, after fusing the weighted behavior codes of job seekers and recruiters into a comprehensive matching code, the matching degree between the comprehensive matching code of job seekers and the feature code of potential matching objects is calculated through a similarity measurement method to determine the best matching result between recruiters and job seekers based on the matching degree. Specifically, the cosine similarity calculation method is used to calculate the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object, and the best matching result between the recruiter and the job seeker is determined based on the matching degree. It can be understood that through the foregoing steps, potential matching objects are intelligently recommended according to the feature data of job seekers and recruiters, improving the matching efficiency and user satisfaction.

[0077] It should be noted that the matching degree refers to the similarity or compatibility degree between the feature vector of a job seeker and the feature vector of a potential matching object. The higher the matching degree score, the better the matching degree between the job seeker and the potential matching object. By finding the best match among multiple recruiters and job seekers, the recruitment efficiency and user satisfaction are improved. And the specific steps for calculating the matching degree are as follows: comparing the comprehensive matching code of the job seeker with the feature code of the potential matching object, and then using the cosine similarity to calculate the matching degree score between the two. That is, by processing the data of job seekers and recruiters from different platforms, a reasonable matching result is given according to the calculation result of the matching degree.

[0078] Among them, calculating the cosine similarity between two vectors, that is:

[0079]

[0080] Among them, the dot product (A, B) represents the dot product of vector A and vector B, the norm(A) represents the Euclidean norm of vector A, and the norm(B) represents the Euclidean norm of vector B.

[0081] Specifically, calculate the comprehensive matching code C comb and the feature code C pot of the potential matching object, that is:

[0082]

[0083] Among them, · represents the dot product of vectors, and ||·|| represents the vector norm.

[0084] It can be seen that in the embodiments of the present invention, the applicant feature vector and the recruiter feature vector corresponding to the applicant and the recruiter are respectively encoded into the applicant behavior code and the recruiter behavior code, and then the feature weights of the applicant and the recruiter are allocated through a dynamic adaptive learning weight allocation mechanism, so that the weights can be dynamically adjusted to optimize the matching effect. Furthermore, through the behavior code fusion strategy, the weighted applicant behavior code and the weighted recruiter behavior code are fused, and then the matching degree between the comprehensive matching code of the applicant obtained by fusion and the feature code of the potential matching object is calculated to determine the best matching result between the recruiter and the applicant based on the matching degree. That is, the information matching implemented by the present application through the dynamic adaptive learning weight allocation mechanism and the behavior code fusion strategy can fully consider and meet the independent preferences of both the applicant and the recruiter, ensure the accuracy of the matching process between the recruiter and the applicant, and thus provide a better service experience for the recruiters and applicants in the recruitment market.

[0085] For example, deep learning frameworks such as TensorFlow and PyTorch can be used to build and train complex recruitment recommendation or matching algorithm models. Machine learning libraries such as scikit-learn provide rich algorithm implementations and are suitable for building recruitment models based on statistical and machine learning methods.

[0086] Please refer to Figure 2 , Figure 2 which is the flowchart of the recruitment system testing method in the present invention. As Figure 2 shown, the recruitment system testing method described in the embodiments of the present invention includes:

[0087] Step S21: Slice the multi-platform recruitment system based on the feature dimension to obtain the recruitment systems corresponding to different slice dimensions for each platform.

[0088] In this embodiment, the different dimensions of the multi-platform recruitment system are sliced to obtain the recruitment systems corresponding to different slice dimensions. That is, for the recruitment system of each platform, it is sliced in the way of dimension slicing to obtain the recruitment systems corresponding to different slice dimensions for each platform.

[0089] Among them, the feature dimension may include user feature dimension, occupation feature dimension, system performance dimension, geographical dimension, and user behavior dimension.

[0090] Step S22: Test and analyze the matching effects corresponding to the recruitment systems of each slice dimension based on the aforementioned information matching method to obtain the test results corresponding to the multi-platform recruitment system.

[0091] In this embodiment, after slicing different dimensions of the multi-platform recruitment system, the matching effects of the recruitment systems under each platform in different dimension slices are respectively tested and analyzed based on the aforementioned information matching method, so that developers can optimize the system performance based on the test results.

[0092] It should be noted that when the matching degree of the recruitment system in a certain dimension slice is generally low, the matching algorithm or feature weight corresponding to this dimension slice needs to be adjusted. Moreover, through stream processing technology, the real-time data in the online recruitment process can be monitored and analyzed, and the test strategy can be adjusted in a timely manner to improve the real-time performance and effectiveness of the test. Then, using distributed storage technology, the test data is scattered and stored on multiple nodes to improve the reliability and availability of the data. At the same time, it is convenient to horizontally and vertically expand the data, or utilize the characteristics of decentralization and immutability of blockchain technology to provide a more secure and transparent data exchange and storage solution for the online recruitment system.

[0093] For the specific content of the aforementioned information matching method, reference can be made to the corresponding content disclosed in the foregoing embodiments, and details will not be elaborated herein.

[0094] It can be seen that in the embodiments of the present application, by slicing different dimensions of the multi-platform recruitment system, the matching effects in different slices are respectively tested and analyzed. By comparing the matching rates and user satisfaction indicators in different slices, the overall performance and optimization direction of the recruitment system are evaluated.

[0095] For example, as shown in Figure 3 In the recruitment system test method, the calculation of the matching degree is achieved through the aforementioned information matching method, that is, the job seeker feature vector and the recruiter feature vector corresponding to the job seeker and the recruiter are respectively encoded into the job seeker behavior code and the recruiter behavior code. Then, through a dynamic adaptive learning weight allocation mechanism, the feature weights of the job seeker and the recruiter are allocated, so that the weights can be dynamically adjusted to optimize the matching effect. Furthermore, through the behavior code fusion strategy, the weighted job seeker behavior code and the weighted recruiter behavior code are fused, and then the matching degree between the comprehensive matching code of the job seeker obtained by fusion and the feature code of the potential matching object is calculated. Then, in the test process, by slicing different dimensions of the multi-platform recruitment system, the matching effects in different slices are respectively tested and analyzed. By comparing the matching rates and user satisfaction indicators in different slices, the overall performance and optimization direction of the recruitment system are evaluated. It is also possible to perform functional testing on the online recruitment system according to the response situations of multiple servers corresponding to multiple clients, and combined with the digital algorithm model to analyze the anomalies and rules of the response data. Moreover, there is information interaction between multiple servers during the online recruitment process, and the digital algorithm model is further used to evaluate the response time and error rate of the server to achieve functional testing and evaluation.

[0096] Among them, multiple clients refer to different user devices or application programs using the online recruitment system, and multiple servers refer to the servers and network infrastructure supporting the online recruitment system, which are responsible for processing client requests, executing business logic, storing data, and providing system functions.

[0097] It should be noted that during the process of testing the recruitment system, the efficiency and accuracy of the testing process can be significantly improved by combining the actual requirements of the online recruitment system testing with digital algorithm models. Among them, the digital algorithm model can include, but is not limited to, machine learning models, deep learning models, and neural network models, which are used for preprocessing of test data, extraction of key features, accurate pattern recognition, and prediction of future results, ensuring a comprehensive and in-depth functional inspection of the online recruitment system.

[0098] For example, the input of the digital algorithm model is the start and end time nodes of each stage of the job application and the corresponding enterprise recruitment evaluation feedback values, and the output is the spatial vector of a single user. Among them, the enterprise recruitment evaluation feedback values can be the evaluations and feedbacks collected from the enterprise during stages such as resume screening, interview, and probation period in the recruitment process, and then these qualitative or quantitative feedbacks can be converted into values that can be used for analysis, such as interview feedback scores, skill matching scores, comprehensive evaluations, recruitment results, job matching degrees, and job seeker performance ratings. Then, based on the spatial vector theory, the job application time, enterprise recruitment evaluation feedback values, and user information are mapped into a three-dimensional space, and the job application status of the user is evaluated by analyzing the relationships between the vectors.

[0099] For example, when analyzing and comparing the job application status of a single user in each stage, the model includes the start time node and end time node of each stage of the job application, which are respectively denoted as Ym-n and Ym, where n = 0, 1, 2, 3, 4...q, n≠q, m = 1, 2, 3, 4,..., p, m≠p, n < m, (n, m ∈ positive real numbers), that is, the range of intervals divided by all job application times is denoted as [Y1], (Y1, Y2), [Y2],..., (Yp-q-1, Yp-1), [Yp-1], (Yp-q, Yp), [Yp]; among them, [Y1] represents the time point of the first stage when the job application starts, (Y1, Y2) represents the time interval from Y1 to Y2, and the representations of other time points and time intervals are the same by analogy.

[0100] The model also includes the enterprise recruitment evaluation feedback values at the start time nodes and end time nodes of each stage of application, denoted as Xb-a and Xb, where a = 0, 1, 2, 3, 4, …, s, a ≠ s, b = 1, 2, 3, 4…h, b ≠ h, a < b, (a, b ∈ positive real numbers), that is, the range corresponding to the single enterprise recruitment evaluation feedback value formed after the interval range division of the learning status evaluation values at all application times is denoted as [X1], (X1, X2), [X2], …, (Xh-s-1, Xh-1), [Xh-1], (Xh-s, Xh), [Xh].

[0101] The model is considered on a single enterprise basis, denoted as Zv, where v = 1, 2, 3, 4, …, u (u belongs to the set of natural numbers N+).

[0102] It can be understood that by considering the start and end time nodes of each stage of recruitment and dividing these time nodes into different interval ranges. At the same time, the feedback values of the enterprise on the recruitment evaluation at all recruitment times are also considered, and these values are divided according to the interval range to form the corresponding range of the single enterprise recruitment evaluation feedback value. Then, by establishing a spatial vector between these time intervals and the range of the enterprise evaluation feedback value and a single user, it can solve the problems in the prior art that, due to the isolated existence of each platform, unified data needs to be repeatedly entered in each module, resulting in the inability of data to be automatically collected, cleaned, sorted, and uniformly stored, and there are problems of data loss and leakage caused by equipment replacement and personnel replacement.

[0103] For example, the interval ranges of the start and end time nodes of each stage of recruitment are expressed as [Y1], (Y1, Y2), [Y2], …, (Yp-q-1, Yp-1), [Yp-1], (Yp-q, Yp), [Yp], and these interval ranges are brought into the Y-axis of the spatial vector; the interval ranges of the recruitment evaluation feedback values of the enterprise at each stage of recruitment time nodes are expressed as [X1], (X1, X2), [X2], …, (Xh-s-1, Xh-1), [Xh-1], (Xh-s, Xh), [Xh], and these interval ranges are brought into the X-axis of the spatial vector; a single user is represented by Zv and is brought into the Z-axis of the spatial vector.

[0104] It should also be noted that the online recruitment system testing method of the present application can be applied to various cloud computing platforms, such as Alibaba Cloud, Tencent Cloud, Huawei Cloud, etc., and the machine learning platforms provided by them, such as Alibaba Cloud PAI (Platform for Artificial Intelligence), Tencent Cloud TI (Tencent Intelligence) platform, or international cloud service providers such as AWS (Amazon Web Services), GCP (Google Cloud Platform), and Azure (Microsoft Azure). Moreover, the present application can use big data processing frameworks such as Hadoop and Spark to process large-scale data in the online recruitment system and provide input for the algorithm model. Stream processing platforms such as Apache Flink can be used to process data information in the online recruitment platform in real time to implement the functions of testing real-time recommendation or matching. Tools such as Swagger and APIGateway (Application Programming Interface Gateway) are used to create and manage API interfaces between the algorithm model and the recruitment system to achieve data exchange and model invocation, and then containerization technologies such as Docker and Kubernetes are used to encapsulate the algorithm model into portable containers for convenient deployment and management of testing in the recruitment system.

[0105] Moreover, for the online recruitment system platform, the present application can be integrated with online recruitment system platforms such as Zhaopin, 51job, and BOSS Zhipin, and the custom algorithm model can be integrated into the system through the APIs or SDKs (Software Development Kits) provided by them.

[0106] In one embodiment, as Figure 4 shown, based on the above information matching method, the present invention also correspondingly provides an information matching system, including:

[0107] A feature determination module 11, configured to obtain independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector;

[0108] A feature encoding module 12, configured to encode the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively;

[0109] A weight determination module 13, configured to determine importance weights of the job seeker behavior encoding and the recruiter behavior encoding in the final matching based on a preset gradient descent algorithm and in an adaptive learning manner, so as to obtain a weighted job seeker behavior encoding and a weighted recruiter behavior encoding;

[0110] A fusion module 14, configured to fuse the weighted job seeker behavior encoding and the weighted recruiter behavior encoding to obtain a comprehensive matching encoding of the job seeker;

[0111] A matching module 15, configured to calculate a matching degree between the comprehensive matching encoding of the job seeker and a feature encoding of a potential matching object, and determine an optimal matching result between the recruiter and the job seeker based on the matching degree.

[0112] Figure 5 It is a schematic structural diagram of a terminal provided in an embodiment of the present application. The terminal may include:

[0113] A memory 501, a processor 502, and a computer program stored on the memory 501 and executable on the processor 502.

[0114] When the processor 502 executes the program, it implements the information matching method provided in the above embodiment.

[0115] Further, the terminal further includes:

[0116] A communication interface 503, configured to communicate between the memory 501 and the processor 502.

[0117] The memory 501 is used to store a computer program executable on the processor 502.

[0118] The memory 501 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0119] If the memory 501, the processor 502, and the communication interface 503 are implemented independently, the communication interface 503, the memory 501, and the processor 502 can be interconnected through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of simplicity in illustration, only one line is used in the figure to represent it, but it does not mean that there is only one bus or one type of bus.

[0120] Optionally, in a specific implementation, if the memory 501, the processor 502, and the communication interface 503 are integrated on a single chip, the memory 501, the processor 502, and the communication interface 503 can communicate with each other through an internal interface.

[0121] The processor 502 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0122] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above information matching method is implemented.

[0123] Those skilled in the art will readily conceive of other implementations of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include known common knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present invention are pointed out by the claims.

[0124] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or N embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0125] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, which can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can read and execute instructions from the instruction execution system, apparatus, or device), or in combination with these instruction execution systems, apparatus, or devices.

[0126] It should be understood that each part of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs, Programmable Gate Arrays), field programmable gate arrays (FPGAs, Field-Programmable Gate Arrays), etc.

[0127] It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or transformations can be made according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. An information matching method, characterized in that: The method comprises: Obtain independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector; Encoding the job seeker feature vector and the recruiter feature vector into job seeker behavior code and recruiter behavior code respectively; Based on a preset gradient descent algorithm, and by means of adaptive learning, the importance weights of the job seeker behavior code and the recruiter behavior code in the final matching are determined to obtain a weighted job seeker behavior code and a weighted recruiter behavior code; The weighted job seeker behavior code and the weighted recruiter behavior code are merged to obtain the comprehensive matching code of the job seeker, and the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object is calculated to determine the best matching result between the recruiter and the job seeker based on the matching degree.

2. The information matching method according to claim 1, characterized in that: The step of acquiring independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector includes: Extracting independent feature data of the job seeker from the detailed features of the job seeker to obtain a feature vector of the job seeker; the detailed features of the job seeker include the job seeker's activity on the recruitment platform, the job seeker's career development, personal information, and job-seeking preferences; Independent feature data of the recruiter is extracted from the detailed features of the recruiter to obtain a recruiter feature vector; the detailed features of the recruiter include basic information of the enterprise, corporate culture and values, historical recruitment records, job requirements, corporate reputation and brand image, financial status and development prospects.

3. The information matching method according to claim 1, characterized in that: The step of encoding the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively comprises: Encoding the job seeker feature vector into a job seeker behavior code using a job seeker coding model constructed based on the first deep neural network model; The recruiter feature vector is encoded into a recruiter behavior code using a recruiter coding model constructed based on the second deep neural network model.

4. The information matching method according to claim 1, characterized in that: The step of fusing the weighted job seeker behavior code and the weighted recruiter behavior code to obtain the job seeker's comprehensive matching code includes: The weighted job seeker behavior code and the weighted recruiter behavior code are merged by using a preset weighted summation method to obtain a comprehensive matching code for the job seeker; Among them, the preset weighted summation method is: C comb =ω j ·C j +ω e ·C e ; Among them, C comb Indicates comprehensive matching code, C j represents the job seeker behavior code, C e represents the recruiter behavior coding, ω j represents the importance weight corresponding to the job seeker behavior coding, ω e Represents the importance weight corresponding to the recruiter's behavior coding.

5. The information matching method according to any one of claims 1 to 4, characterized in that: The calculating the matching degree between the comprehensive matching code of the job seeker and the characteristic code of the potential matching object comprises: The cosine similarity calculation method is used to calculate the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object.

6. The information matching method according to claim 5, characterized in that: Also includes: Extracting English application ability feature vector from the English test results of the job applicant; Determine the importance weight of the English application ability feature vector in the final matching based on the reinforcement learning algorithm and the gradient descent algorithm to obtain a weighted English application ability feature vector; The step of fusing the weighted job seeker behavior code and the weighted recruiter behavior code to obtain the job seeker's comprehensive matching code includes: The weighted job seeker behavior code, the weighted English application ability feature vector and the weighted recruiter behavior code are integrated to obtain the comprehensive matching code of the job seeker.

7. A recruitment system testing method, characterized in that: The method comprises: Slice the multi-platform recruitment systems based on feature dimensions to obtain recruitment systems with different slicing dimensions corresponding to each platform; Based on the information matching method described in any one of claims 1 to 6, the matching effect corresponding to the recruitment system of each slice dimension is tested and analyzed respectively to obtain the test results corresponding to the multi-platform recruitment system.

8. An information matching system, characterized in that: The system comprises: A feature determination module is used to obtain independent feature data of job seekers and recruiters to obtain a job seeker feature vector and a recruiter feature vector; A feature encoding module, used for encoding the job seeker feature vector and the recruiter feature vector into a job seeker behavior code and a recruiter behavior code respectively; A weight determination module, for determining the importance weights of the job seeker behavior code and the recruiter behavior code in the final matching based on a preset gradient descent algorithm and by adaptive learning, to obtain a weighted job seeker behavior code and a weighted recruiter behavior code; A fusion module, used for fusing the weighted job seeker behavior code and the weighted recruiter behavior code to obtain a comprehensive matching code for the job seeker; The matching module is used to calculate the matching degree between the comprehensive matching code of the job seeker and the feature code of the potential matching object to determine the best matching result between the recruiter and the job seeker based on the matching degree.

9. A terminal, characterized in that: include: A memory, a processor, and an information matching program stored in the memory and executable on the processor, wherein the information matching program, when executed by the processor, implements the steps of the information matching method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which can be executed to implement the steps of the information matching method according to any one of claims 1 to 6.

Images