An intelligent matching method and system for job resumes based on multi-dimensional analysis

Abstract

The present invention relates to the technical field of human resource information processing, and specifically relates to an intelligent job resume matching method and system based on multi-dimensional analysis, including a data collection module, a data processing module, and a data matching module. The data collection module is used to obtain resume information data of job seekers and recruitment information data of enterprises. The data processing module is used to extract information dimensions from the resume information data and the recruitment information data. By quantitatively calculating the scores of all information dimensions in the hard information data and the soft information data, and combining them to calculate the matching score, the present invention can more comprehensively evaluate the matching degree between job seekers and positions, automatically identify the demand differences of job seekers according to the importance degree of job seekers for information dimensions, and then update the weight coefficients to accurately meet the personalized needs of job seekers, provide accurate matching of rigid requirements for job seekers and enterprises, and also provide flexible alternative options for job seekers and enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of human resource information processing, and particularly relates to a method and system for intelligent matching of job resumes based on multi-dimensional analysis. Background Art

[0002] With the rapid development of global technology and the continuous growth of human resource management needs, the traditional recruitment model can no longer meet the requirements of modern enterprises for efficiency and accuracy. The massive enterprise job demands and job seeker resume data have made the tasks of the human resources department in the recruitment process heavy and inefficient, and simply relying on manual matching of jobs and resumes can no longer meet the timeliness and scientific needs of modern enterprises. Against this background, an intelligent job resume matching system based on multi-dimensional analysis has emerged. Such a system constructs an intelligent and automated talent screening and matching process through technical means and has become an important tool for improving recruitment efficiency in the information age.

[0003] After retrieval, Chinese Patent No. CN201610784547.4 discloses a method for matching a resume with a job, which includes: obtaining the behavioral attribute information of a job seeker user and the behavioral attribute information of a recruiter user according to the access records of the job seeker user and the recruiter user on the recruitment platform; determining a first matching score between the resume of the job seeker user and the job of the recruiter user according to the resume of the job seeker user and the job of the recruiter user; determining a second matching score between the resume of the job seeker user and the job of the recruiter user according to the behavioral attribute information of the job seeker user and the recruiter user; and determining the total matching score between the resume of the job seeker user and the job of the recruiter user according to the first matching score and the second matching score.

[0004] The above solution performs matching analysis and processing on the resume of the job seeker user and the job of the recruiter user, and also performs matching analysis and processing on the eagerness degree of both parties according to the behavioral attribute information of both parties. By performing various matching degree analyses on the resume and the job, considering various factors for comprehensive ranking, and automatically generating the optimal matching set of the resume and the job, it realizes optimal matching and accurate delivery, can greatly improve the interview conversion rate of the recruitment platform's proxy submission, enables job seekers to find the desired job interview opportunities faster and more accurately, and enables recruitment enterprises to find ideal talents more timely and accurately. Correspondingly, it can increase the revenue of the recruitment platform.

[0005] However, most current job resume matching systems rely on keyword template matching or shallow text similarity calculation. The analysis methods are single, and it is easy to ignore the context semantics and detailed information, resulting in inaccurate matching. For example, most job resume matching systems only focus on the matching of basic hard conditions, such as work experience, education background, skill lists, etc., and ignore soft conditions, such as work preferences, career interests, cultural values, etc. In addition, the data in dimensions such as soft conditions are relatively subjective, and it is difficult for existing job resume matching systems to accurately extract their quantitative data, which further leads to the difficulty of the job resume matching system in comprehensively supporting the needs of job seekers and recruiting enterprises.

[0006] Therefore, a job resume intelligent matching method and system based on multi-dimensional analysis are proposed to solve the above problems. Summary of the Invention

[0007] Technical Problems to be Solved

[0008] In view of the above-mentioned shortcomings of the prior art, the present invention provides a job resume intelligent matching method and system based on multi-dimensional analysis, which can effectively solve the problem that most job resume matching systems in the prior art rely on basic hard condition data for matching while ignoring the consideration of soft conditions, and at the same time cannot accurately extract and quantify the subjective data in soft conditions.

[0009] Technical Solutions

[0010] To achieve the above object, the present invention is realized through the following technical solutions:

[0011] The present invention provides a job resume intelligent matching system based on multi-dimensional analysis, including a data acquisition module, a data processing module, and a data matching module. The data acquisition module is used to obtain the resume information data of job seekers and the recruitment information data of enterprises. The data processing module is used to extract the information dimensions in the resume information data and the recruitment information data, calculate the job seeker suitability score and the enterprise suitability score and output them. The data matching module is based on the job seeker suitability score and the enterprise suitability score to calculate the matching score between the job seeker and the enterprise , and use the matching score to judge whether to match the resume information data of the job seeker and the recruitment information data of the enterprise; an optimization model is constructed based on the information dimension to obtain an optimized matching score ; the data matching module constructs a matching matrix M based on the optimized matching score , and inputs the matching matrix M into the matching model to obtain the globally optimal matching result.

[0012] Further, both the resume information data and the recruitment information data include hard information data and soft information data; among them, the information dimensions of the hard information data include: educational background information, work information, skill information, and requirement information; the information dimensions of the soft information data include: career interest, career direction, cultural fit, and career stability.

[0013] Further, the method for calculating the job seeker's fitness score and the enterprise's fitness score includes: defining the information dimension scores in the hard information data and the soft information data; based on the information dimension scores, calculating the job seeker's hard score , the job seeker's soft score , the enterprise's hard score and the enterprise's soft score , and the calculation formulas are:

[0014] ;

[0015] In the formula, is the job seeker's educational background score, is the job seeker's work score, is the job seeker's skill score, is the job seeker's requirement score, is the influence coefficient of the corresponding information dimension;

[0016] ;

[0017] In the formula, is the job seeker's interest score, is the job seeker's direction score, is the job seeker's cultural score, is the job seeker's stability score, is the influence coefficient of the corresponding information dimension;

[0018] ;

[0019] In the formula, is the enterprise's educational background score, is the enterprise's work score, is the enterprise's skill score, is the enterprise's requirement score;

[0020] ;

[0021] In the formula, is the enterprise's interest score, is the enterprise's direction score, is the enterprise's cultural score, is the enterprise's stability score;

[0022] ;

[0023] ;

[0024] In the formula, and are the weight coefficients of the hard information data and the soft information data respectively.

[0025] Furthermore, the formula for calculating the matching score between the job seeker and the enterprise is: : ; In the formula, and are the weight coefficients of the job seeker's fitness score and the enterprise's fitness score respectively. The matching score is used to compare with the pre-set threshold score .

[0026] Furthermore, the method for constructing the optimization model includes: constructing a feature vector according to the information dimension score, performing normalization processing on the information dimension score, mapping all eigenvalues to the interval [0,1] to obtain the hard information feature vector and the soft information feature vector ; Using the hard information feature vector and the soft information feature vector to construct a comprehensive feature vector ;

[0027] Calculate the mathematical distance between the job seeker's feature vector and the position feature vector. The calculation formula is:

[0028] Distance ; In the formula, represents the job seeker's feature vector, ; represents the position feature vector, ; n is the number of features in the vector; , is the value of the i-th eigenvalue, belonging to the i-th dimension of the job seeker's feature vector and the position feature vector respectively;

[0029] Calculate the distance between each sample and its target sample , and record the distance in order and the corresponding sample index i; According to the calculated distance d value, sort all samples in ascending order to obtain a sorted list ; In the formula, is the sorted distance; is the smallest sorted distance, and the corresponding sample is the sample closest to the target sample; is the largest sorted distance, and the corresponding sample is the sample farthest from the target sample; Select the top K samples with the smallest distances from the sorted list List to construct the nearest neighbor set ; Assign weights to each nearest neighbor sample , and the calculation formula is: ; In the formula, is the distance of the i-th neighbor;

[0030] Optimize the job seeker fitness score and the enterprise fitness score , and the calculation formula is:

[0031] ; In the formula, is the optimized job seeker fitness score; represents the job seeker fitness score of the i-th neighbor;

[0032] ; In the formula, is the optimized enterprise fitness score; represents the enterprise fitness score of the i-th neighbor;

[0033] Use the optimized job seeker fitness score and the optimized enterprise fitness score to calculate the optimized matching score between the job seeker and the enterprise .

[0034] Furthermore, the construction method of the matching model includes:

[0035] Define the format of the matching matrix M: The rows of the matching matrix M represent each job seeker ; The columns of the matching matrix M represent each position ; The element of the matching matrix M represents the optimized matching score between the job seeker and the position ; Based on the magnitudes of the elements in each row of the matching matrix M, determine the position priority list of each job seeker ; Based on the magnitudes of the elements in each column of the matching matrix M, determine the job seeker priority list of each position ;

[0036] Set the input as the job priority list and the job seeker priority list ; Set the output as the optimal matching result;

[0037] Initialize the matching status and perform step-by-step matching. Repeat the step-by-step matching operation until all job seekers are all matched or all jobs are all occupied, and all job seekers and jobs no longer have better options; Output the optimal matching result without blocking pairs.

[0038] Furthermore, the data matching module constructs a feedback model based on the dynamic behaviors and selection preferences of job seekers, generates feedback information for updating the weight coefficients of hard information data and the weight coefficients of soft information data , and obtain the updated weight coefficients of hard information data that meet the constraint conditions and the updated weight coefficients of soft information data ; Based on the updated weight coefficients of hard information data and the updated weight coefficients of soft information data calculate the updated job seeker fitness score and the updated enterprise fitness score , calculate the updated matching score , and return it to the matching model to obtain a new optimal matching result;

[0039] The feedback information includes: the information dimension scores of the jobs where the job seekers failed to match successfully in the first round, the secondary jobs automatically assigned, and the new jobs reselected; The constraint condition is .

[0040] Furthermore, the method for updating the weight coefficients of hard information data and the weight coefficients of soft information data includes:

[0041] ; Wherein, is the learning rate; k is the index of the information dimension; represents the sum of the absolute values of the differences of the information dimension scores in the hard information data; n is the total number of information dimensions in the hard information data; is the difference value of the information dimension, and the calculation formula is: ; Wherein, is the information dimension score of the secondary job, is the dimension score of the new job;

[0042] ; wherein, represents the sum of the absolute values of the differences in the information dimension scores in the soft information data; m is the total number of information dimensions in the soft information data.

[0043] Furthermore, the data matching module inputs into the weight dynamic adjustment model and compares it with a preset threshold T; if , the weight of the information dimension with index k is increased; if , the adjustment of the weight of the information dimension with index k is ignored.

[0044] An intelligent matching method for job resumes based on multi-dimensional analysis, comprising the following steps:

[0045] Step 1: Collect the resume information data of job seekers and the recruitment information data of enterprises, and extract the information dimensions therein;

[0046] Step 2: Calculate the job seeker fitness score and the enterprise fitness score based on the information dimensions;

[0047] Step 3: Calculate the matching score between the job seeker and the enterprise based on the job seeker fitness score and the enterprise fitness score, compare the matching score with a preset threshold score, and preferentially match the job seekers and enterprises whose matching scores are greater than or equal to the threshold score.

[0048] Beneficial effects

[0049] The technical solution provided by the present invention has the following beneficial effects compared with the prior art:

[0050] This solution can more comprehensively evaluate the matching degree between job seekers and positions by quantitatively calculating all information dimension scores in the hard information data and the soft information data and combining them to form the job seeker fitness score and the enterprise fitness score ; the hard information data can evaluate the basic qualifications of job seekers, and the soft information data can evaluate the potential fit of job seekers. When evaluating job seekers, if the hard information data of a job seeker fully meets the position requirements but the soft information data score is low, the final comprehensive score will be appropriately reduced, thus avoiding the risk of recruitment failure. On the contrary, if a job seeker is slightly lower than the position requirements in terms of hard information data but has a high soft information score, the comprehensive score will be appropriately increased, thereby providing opportunities for job seekers;

[0051] Through the settings of the feedback model and the weight dynamic adjustment model, this solution can more accurately meet the two-way matching needs of job seekers and enterprises; during the matching process, according to the importance degree of job seekers for hard information data and soft information data, it automatically identifies the demand differences of job seekers, and then the weight coefficient of hard information data and the weight coefficient of soft information data are updated to accurately meet the personalized needs of job seekers, that is, it can provide accurate matching rigid requirements for job seekers and enterprises, and can also provide flexible alternative solutions for job seekers and enterprises; at the same time, it gives feedback on each failed match between job seekers and enterprises, records the behavior data of job seekers, and then captures the demand preferences of job seekers, improves the self-adaptability of the matching model, realizes the gradual optimization of the matching effect, makes the matching logic more flexible and adaptable, and thus greatly improves the matching success rate. Brief Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained according to these drawings without creative efforts.

[0053] Figure 1 It is a schematic diagram of the composition of the intelligent job resume matching system in Embodiment 1 of the present invention.

[0054] Figure 2 It is a schematic diagram of the flow of the intelligent job resume matching method in Embodiment 2 of the present invention. Detailed Embodiment

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0056] The following further describes the present invention with reference to the embodiments.

[0057] Embodiment 1:

[0058] Refer to the appendix Figure 1, this case proposes an intelligent job resume matching system based on multi-dimensional analysis, including a data collection module, a data processing module, and a data matching module; among them, the data collection module is used to obtain the resume information data of job seekers and the recruitment information data of enterprises; the data processing module is used to extract the information dimensions in the resume information data and the recruitment information data, and calculate their respective fitness scores (the job seeker fitness score and the enterprise fitness score ) and then output; the data matching module is used to receive the fitness scores output by the data processing module, calculate the matching score of the job seeker and the enterprise, compare it with the pre-set threshold score, and judge whether to match the resume information data of the job seeker and the recruitment information data of the enterprise, so as to provide the best two-way choice for the job seeker and the enterprise.

[0059] Specifically, both the resume information data and the recruitment information data include hard information data and soft information data; among them, the information dimensions of the hard information data include: educational information (i.e., the graduate school, degree, major, graduation time, etc. of the job seeker himself and the educational and major requirements of the enterprise for the job seeker), work information (i.e., the working years of the job seeker, relevant industry experience, and achievements obtained during the work, etc., and the work requirements of the enterprise for the job seeker), skill information (i.e., the professional skills mastered by the job seeker, the professional qualification certificates obtained, and the skill requirements of the enterprise for whether the job seeker has the position qualifications), demand information (i.e., the salary expectation and target work location of the job seeker, and the job form information provided by the enterprise); the information dimensions of the soft information data include: career interest (i.e., the degree of interest of the job seeker himself in the position and the attractiveness of the recruitment industry and position to the job seeker), career direction (i.e., the career development plan of the job seeker himself and the development concept of the enterprise for the position), cultural fit (i.e., whether the job seeker can adapt to and integrate into the enterprise work culture), career stability (i.e., whether the job position of the enterprise has long-term development, or whether the job seeker has the stability of long-term work).

[0060] Furthermore, in this solution, the method of calculating the job seeker fitness score and the enterprise fitness score based on the information dimensions in the resume information data and the recruitment information data includes:

[0061] Define the information dimension scores in the hard information data, which are respectively the educational score (including the job seeker educational score and the enterprise educational score ), the work score (including the job seeker work score and the enterprise work score ), the skill score (including the skill scores of job seekers and the skill scores of enterprises ) and the demand scores (including the demand scores of job seekers and the demand scores of enterprises );

[0062] Define the information dimension scores in the soft information data, namely the interest scores (including the interest scores of job seekers and the interest scores of enterprises ), the direction scores (including the direction scores of job seekers and the direction scores of enterprises ), the culture scores (including the culture scores of job seekers and the culture scores of enterprises ) and the stability scores (including the stability scores of job seekers and the stability scores of enterprises );

[0063] Based on the information dimension scores, calculate the hard scores of job seekers and the soft scores of job seekers , and the calculation formula is:

[0064] ;

[0065] In the formula, is the influence coefficient of the corresponding information dimension, reflecting the relative importance of each information dimension to the overall matching score ;

[0066] ;

[0067] In the formula, is the influence coefficient of the corresponding information dimension, reflecting the relative importance of each information dimension to the overall matching score ;

[0068] Based on the hard scores of job seekers and the soft scores of job seekers calculate the fitness scores of job seekers , and the calculation formula is: ;

[0069] In the formula, and are the weight coefficients of the hard information data and the soft information data respectively, reflecting the relative importance of each information dimension to the overall matching score .

[0070] Calculate the hard score of the enterprise based on the information dimension score and the soft score of the enterprise , and the calculation formula is:

[0071] ;

[0072] ;

[0073] Based on the hard score of the enterprise and the soft score of the enterprise calculate the enterprise fitness score , and the calculation formula is ;

[0074] Based on the above, this solution provides an operable basis for the precise matching of job seekers and enterprises through the mechanical quantitative calculation and analysis of the information dimensions in the resume information data and recruitment information data.

[0075] It should be noted that the hard score of the job seeker and the hard score of the enterprise are usually quantified and regularized, and can be calculated by explicit quantification or assignment according to the matching degree. Specifically:

[0076] The calculation method of the education score is as follows:

[0077] For the education requirement, several levels are set. For example, if the job seeker's education is higher than the minimum requirement of the position, the assigned value is 1; if the job seeker's education is equal to the minimum requirement of the position, the assigned value is 0.8; if the job seeker's education requirement is lower than but close to the position requirement, the negative value is 0.4; if the job seeker's education requirement completely does not match the position requirement, the negative value is 0;

[0078] For the matching degree of professionalism, several levels are set. For example, if the job seeker's major completely matches the position requirement, the assigned value is 1; if the job seeker's major partially matches the position requirement (such as belonging to the same major category, for example, computer technology and software engineering), the assigned value is 0.8; if the job seeker's major does not match but is related to the position requirement (belonging to adjacent disciplinary fields or having a certain intersection, for example, computer technology and mathematics), the assigned value is 0.4; if the job seeker's major completely does not match the position requirement, the assigned value is 0;

[0079] For the assigned values of institutions, several levels are also set. For example, for top institutions, the assigned value is 1; for key undergraduate institutions, the assigned value is 0.7; for ordinary undergraduate institutions, the assigned value is 0.5; for junior colleges, the assigned value is 0.1.

[0080] .

[0081] Work score The calculation method is as follows:

[0082] ; In the formula, is the industry relevance score, which is used to measure whether the job seeker's work experience is relevant to the industry to which the target position belongs; is the work achievement score, which is used to measure whether the job seeker has made significant actual performance during the work; , , are the weight coefficients of the corresponding parameters; The industry relevance score and the work achievement score are both assigned according to the matching degree.

[0083] Skill score The calculation method is as follows: ;

[0084] Requirement score The calculation method is as follows: ; In the formula, is the salary score, and its calculation formula is , by calculating the ratio of the salary offered by the enterprise and the expected salary of the job seeker and quantifying the salary matching score , if the ratio is greater than 1, it means that the salary level of the enterprise fully meets the target expectation of the job seeker and even provides additional attraction. If the ratio is less than 1 and the smaller the ratio, it means that the gap between the salary level of the enterprise and the job seeker is larger; is the work location score, and its calculation formula is , where d is the actual distance between the address of the job seeker and the enterprise office location, is the maximum acceptable distance threshold set by the job seeker, is to ensure that the calculation result is not negative and avoid unreasonable negative scores; is the work form score, which is used to match whether the job seeker accepts the work mode provided by the enterprise and is also assigned according to the matching degree; , , are the weight coefficients of the corresponding parameters.

[0085] Job seeker soft score and enterprise soft score are calculated based on the information dimensions in the soft information data, and the information dimensions in the soft information data are dimensions related to the information dimensions in the hard information data. Usually, they cannot be directly quantified, but by analyzing the content of the user's resume, behavior data or feedback, a quantitative model is constructed based on the analyzed data to capture these information dimensions and map them into the standardized framework of the system, and then quantified calculation. Specifically:

[0086] They are respectively: interest score The calculation method is as follows:

[0087] ; In the formula, is a similarity function, which is used to calculate the semantic similarity between the job-hunting keywords in the resume information data and the position keywords in the recruitment information data (for example, by converting the job-hunting keywords and position keywords into word vectors and using cosine similarity to calculate the semantic similarity between the job-hunting keywords and position keywords). The score range is [0, 1]. If , it means that the job-hunting keywords and position keywords are completely matched (i.e., the semantics are the same). If , it means that the job-hunting keywords and position keywords are partially matched (i.e., including near-synonym relationships or relevance). If , it means that the job-hunting keywords and position keywords are completely unmatched; n is the number of pairs of matched job-hunting keywords and position keywords; and then the interest score is obtained by quantification;

[0088] Direction score The calculation method is as follows: ; By extracting the description of the career goal from the resume information data of the job seeker and extracting the position development path from the recruitment information data, and then using the similarity function to calculate the similarity between the career goal of the job seeker and the position development path of the recruitment position; and then the direction score is obtained by quantification;

[0089] Cultural score The calculation method is: ; By extracting the description of the personality keywords from the resume information data of the job seeker and extracting the cultural keywords of the enterprise from the recruitment information data, and then using the similarity function to calculate the similarity between the personality keywords of the job seeker and the cultural keywords of the enterprise; and then the cultural score is obtained by quantification;

[0090] Stability score The calculation method is as follows: ; In the formula, is the average duration of each job of the job seeker; is a scoring function, which is used to map the average duration to the stability score . In this case, the scoring function can be set as a piecewise function, such as: , by setting the average duration into three intervals and mapping different intervals to different values, and then the stability score is obtained by quantification.

[0091] In summary, by calculating the scores of all information dimensions in the hard information data and soft information data and combining them to form the candidate suitability score and the enterprise suitability score , the matching degree between the candidate and the position can be evaluated more comprehensively; the hard information data can evaluate the basic qualifications of the candidate, and the soft information data can evaluate the potential fit of the candidate. When evaluating a candidate, if the hard information data of the candidate fully meets the position requirements but the score of the soft information data is low, the final comprehensive score will be appropriately reduced to avoid the risk of recruitment failure. On the contrary, if the candidate is slightly lower than the position requirements in the hard information data but has a high score in the soft information, the comprehensive score will be appropriately increased to provide an opportunity for the candidate.

[0092] The data matching module calculates the matching scores of the candidate and the enterprise in the following way:

[0093] ; and are the weight coefficients of the candidate suitability score and the enterprise suitability score respectively.

[0094] Define the threshold scores as high matching degree and low matching degree respectively. Compare the matching score with the threshold score :

[0095] If , it means that the candidate and the enterprise position are highly matched in both hard information data and soft information data, and it is suitable for priority recommendation;

[0096] If , it means that there is a certain degree of matching between the candidate and the enterprise position, but some dimensions do not reach the optimal standard, and it is suitable for recommending to candidates or enterprise positions in specific scenarios, and further communication between the candidate and the enterprise is required to determine;

[0097] If , it means that the candidate and the enterprise position do not match in hard information data and soft information data seriously, and it is not recommended for recommendation. It is suitable for excluding inappropriate candidates or positions to improve the screening efficiency;

[0098] Finally, recommend the candidates and enterprises with to each other to provide the best two-way choices for candidates and enterprises.

[0099] The data processing module is performing the matching score of the candidate and the enterprise When performing calculations, the KNN algorithm is defined as an optimized matching score The optimized model. According to the information dimension scores of job seekers and enterprises, construct feature vectors, normalize the information dimension scores, map all eigenvalues to the interval [0, 1], and obtain the hard information feature vector And the soft information feature vector ; Then use the hard information feature vector And the soft information feature vector To construct a comprehensive feature vector ;

[0100] Define the Euclidean distance as the measurement method for job seekers and enterprises, calculate the mathematical distance between the job seeker's feature vector and the position feature vector (i.e., the degree of proximity between the job seeker and the enterprise in the feature space). The smaller the distance between the two, the closer they are; the larger the distance, the greater the difference in their features, that is, the lower the matching degree; the calculation formula is: distance ; In the formula, Represents the job seeker's feature vector, ; Represents the position feature vector, ; n is the number of features in the vector, that is, the dimension of the feature vector, representing the number of elements participating in the similarity calculation; , Is the value of the i-th eigenvalue, belonging to the i-th dimension of the job seeker's feature vector And the position feature vector Respectively;

[0101] For all job seekers And positions , calculate the distance Of each sample And its target sample , and record the distance In order and the corresponding sample index i; According to the calculated distance d value, sort all samples in ascending order to obtain a sorted list ; In the formula, Is the sorted distance; Is the smallest distance after sorting, and the corresponding sample Is the sample closest to the target sample ; Is the largest distance after sorting, and the corresponding sample Is the sample farthest from the target sample ; Through sorting, the sample with the smallest distance is arranged at the beginning of the sorted list List, and the largest distance is arranged at the end; Select the first K samples (neighbors) with the smallest distance from the sorted list List to construct the nearest neighbor set Assign weights to each nearest neighbor sample , reflecting the relationship between distance and matching score , and the calculation formula is as follows: ; In the formula, is the distance of the i-th neighbor; From the above formula, it can be seen that the smaller the distance of the sample, the higher the weight, because they are closer and have greater reference value;

[0102] Use the scores of the K neighbors found by the optimization model to optimize the candidate fitness score and the enterprise fitness score , and the calculation formula is as follows:

[0103] ; In the formula, is the optimized candidate fitness score, indicating that after combining the score information of K neighbors, a more reasonable comprehensive score is obtained; represents the candidate fitness score of the i-th neighbor;

[0104] ; In the formula, is the optimized enterprise fitness score, indicating that after combining the score information of K neighbors, a more reasonable comprehensive score is obtained; represents the enterprise fitness score of the i-th neighbor;

[0105] Use the optimized candidate fitness score and the optimized enterprise fitness score to calculate the optimized matching score between the candidate and the enterprise .

[0106] Furthermore, in this case, not only the optimized matching score between a single candidate and a single enterprise position needs to be calculated , but also consider that a candidate may be suitable for different positions of multiple enterprises, and the positions of an enterprise will also be suitable for multiple different candidates. Therefore, it is necessary to find the combination with the highest fitness among all candidates and enterprise positions to meet the needs of both parties as much as possible. The data matching module constructs a matching matrix M based on the optimized matching scores of each candidate and each position. The rows of the matching matrix M represent each candidate (such as a total of n candidates), and the columns of the matching matrix M represent each position (such as a total of m positions). The element of the matching matrix M represents the optimized matching score between candidate and position , and the range is [0, 1]. The higher the value, the higher the matching degree.

[0107] Based on the magnitudes of the elements in each row of the matching matrix M to determine the job priority list for each job seeker ; based on the magnitudes of the elements in each column of the matching matrix M to determine the job seeker priority list for each job ; input the matching matrix M into a pre - constructed matching model to obtain the globally optimal matching result; the construction method of the matching model is as follows: Set the input (i.e., the job priority list and the job seeker priority list

[0108] ) and the output (the optimal matching result); then initialize the matching status, that is, all job seekers are in the unmatched state and all jobs are in the vacant state; then perform step - by - step matching, that is, each job seeker submits applications to the jobs in its job priority list in sequence. If there is no matching object for the job in the job priority list , then accept the current application, that is, the job seeker matches the job ; if the job already has a matched job seeker (denoted as ), but the newly applying job seeker (job seeker ) has a higher priority, that is, the optimized matching score of the job seeker is greater than the optimized matching score of the job seeker , then the job is replaced by the new job seeker, that is, the job matches the job seeker ; at the same time, reject the original matching object and make it return to the free state, that is, the job seeker becomes unmatched; repeat the above steps to make the rejected job seeker continue to submit applications to the next job in its job priority list and keep operating until all job seekers are matched or all jobs are occupied, and there are no better options for all job seekers and jobs ; the finally output optimal matching result satisfies stability (i.e., there are no blocking pairs).

[0109] ​​​Example: Assume there are three job seekers and three positions:

[0110] Job seeker 's job preference list ;

[0111] Job seeker 's job preference list ;

[0112] Job seeker 's job preference list ;

[0113] Position 's job seeker preference list ;

[0114] Position 's job seeker preference list ;

[0115] Position 's job seeker preference list ;

[0116] Initialize the matching status, i.e., job seekers , , are all in a free state, and positions , , are all vacant; start the step-by-step matching. Round 1:

[0117] Job seeker applies to position , and position accepts job seeker ;

[0118] Job seeker also applies to position , and position compares job seeker and job seeker 's priorities and still selects job seeker , rejecting job seeker ;

[0119] Job seeker applies to position again, and position accepts job seeker ;

[0120] Job seeker applies to position , and position compares job seeker and job seeker Prioritize and select job seekers , reject job seekers ;

[0121] At this time, the matching status is: job seeker matches the position , job seeker matches the position , job seeker is in a free state; start round two:

[0122] Job seeker also applies to the position , the position accepts the job seeker ;

[0123] At this time, the matching status is: job seeker matches the position , job seeker matches the position , job seeker matches the position ; Matching ends, all job seekers have been matched, and the results are stable.

[0124] Through the setting of the matching model, the interactive mode of job seekers taking the initiative to initiate matching and positions responding is iterated repeatedly until the matching is stable, which can ensure the stability of the two-way matching between job seekers and enterprise positions. At the same time, considering the priority ranking of job seekers and enterprise positions, as fair an allocation as possible is achieved during the two-way matching process.

[0125] It is worth mentioning that when the matching model completes the matching of job seekers and enterprise positions, although it can solve the basic stability problem, during the multi-round matching process, some job seekers may still have incomplete matching problems due to information loss or weight deviation during matching; for example, a job seeker does not enter the position with the highest matching degree during the initial round of matching, and after matching with a secondary matching degree position, the reverse position is not suitable and needs to reapply for the first-level position; during this process, the data matching module constructs a feedback model based on the dynamic behavior and selection preferences of job seekers and generates feedback information to optimize the weight coefficients of hard information data and soft information data , to improve the success rate and accuracy of subsequent matching between job seekers and positions.

[0126] The way to construct the feedback model is as follows:

[0127] Collect the behavioral records of job seekers, including: the primary positions that the job seekers did not match successfully in the first round (i.e., the positions with the highest matching degree), the secondary positions that the job seekers were assigned in the second-round matching (i.e., the positions with lower matching degree), and the new positions that the job seekers re-applied for; collect the information dimension scores of the primary positions and the secondary positions; and collect the difference value DV between the information dimension scores. Based on the difference value DV, update the weight coefficients of the hard information data and the weight coefficients of the soft information data to obtain the updated weight coefficients of the hard information data and the updated weight coefficients of the soft information data , then calculate the updated job seeker fitness score and the updated enterprise fitness score , and then use the updated job seeker fitness score and the updated enterprise fitness score to calculate the updated matching score , and return the updated matching score to the matching model, and then update the matching of the job seekers and the enterprise positions to obtain a new optimal matching result.

[0128] The method for updating the weight coefficients of the hard information data and the weight coefficients of the soft information data is as follows:

[0129] The difference value ; where is the information dimension score of the secondary position, is the dimension score of the new position;

[0130] ; where is the updated weight coefficient of the hard information data; is the learning rate; k is the index of the information dimension, representing a specific information dimension; represents the sum of the absolute values of the difference values of the information dimension scores in the hard information data, reflecting the overall matching deviation of the hard information data; n is the total number of information dimensions in the hard information data, used for normalization calculation;

[0131] ; where is the updated weight coefficient of the soft information data; represents the sum of the absolute values of the difference values of the information dimension scores in the soft information data, reflecting the overall matching deviation of the soft information data; m is the total number of information dimensions in the soft information data;

[0132] At the same time, add a constraint condition .

[0133] After calculating the difference values of different information dimensions the difference values are input into a pre-constructed weight dynamic adjustment model, and the difference values are compared with a pre-set threshold T; if , it means that when the matching model matches job seekers and enterprises, the actual needs deviation of the job seeker is relatively large. Therefore, it is necessary to increase the weight of the information dimension with index k; for example , it is necessary to increase the weight coefficient of the updated hard information data to improve the influence of the updated job seeker fitness score on the updated matching score ; if , it means that the matching of the matching module in the information dimension with index k is accurate enough and there is no obvious deviation. Therefore, the weight adjustment can be turned to other dimension information to optimize the overall matching.

[0134] In summary, through the settings of the feedback model and the weight dynamic adjustment model, this solution can more accurately meet the two-way matching needs of job seekers and enterprises; in the matching process, according to the importance degree of job seekers for hard information data and soft information data, the demand differences of job seekers are automatically identified, and then the weight coefficient of the hard information data and the weight coefficient of the soft information data are updated to accurately meet the personalized needs of job seekers, that is, it can provide accurate matching rigid requirements for job seekers and enterprises, and can also provide flexible alternative solutions for job seekers and enterprises; at the same time, feedback is given for each failed match between job seekers and enterprises, and the behavior data of job seekers is recorded, so as to capture the demand preferences of job seekers, improve the self-adaptability of the matching model, realize the gradual optimization of the matching effect, make the matching logic more flexible and adaptable, and thus greatly improve the matching success rate.

[0135] Embodiment 2:

[0136] Refer to the appendix Figure 2 , on the basis of Embodiment 1, this case proposes an intelligent matching method for job resumes based on multi-dimensional analysis, including the following steps:

[0137] Step 1: Collect the resume information data of job seekers and the recruitment information data of enterprises, and extract the information dimensions therein;

[0138] Step 2: Calculate the job seeker fitness score and the enterprise fitness score based on the information dimensions;

[0139] Step 3: Calculate the matching scores of job seekers and enterprises based on the job seeker fitness scores and enterprise fitness scores, compare the matching scores with the pre-set threshold scores, and preferentially match the job seekers and enterprises whose matching scores are greater than or equal to the threshold scores.

[0140] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A job resume intelligent matching system based on multi-dimensional analysis, including a data collection module, a data processing module and a data matching module. The data collection module is used to obtain the resume information data of job seekers and the recruitment information data of enterprises, and is characterized in that: The data processing module is used to extract information dimensions from resume information data and recruitment information data, and calculate the job seeker's suitability score based on the information dimensions. And enterprise adaptability And output, the data matching module is based on the job seeker's suitability score And enterprise adaptability Calculate the matching score between job seekers and companies , using matching score Determine whether to match the resume information data of the job seeker with the recruitment information data of the enterprise; Build an optimization model based on information dimensions to obtain optimized matching scores ; The data matching module is based on the optimized matching score Construct a matching matrix M, input the matching matrix M into the matching model, and obtain the global optimal matching result; The calculation of the job applicant's suitability score And enterprise adaptability The methods include: defining information dimension scores in hard information data and soft information data; calculating the hard scores of job seekers based on the information dimension scores , Soft Rating of Job Applicants , Enterprise hard scoring and enterprise soft rating ; ； ； In the formula, and are the weight coefficients of hard information data and soft information data respectively; Calculate the matching score between job seekers and enterprises The formula is: ; In the formula, and The job seeker suitability scores are And enterprise adaptability Weight coefficient, matching score Used to compare with pre-set thresholds Make a comparison; The data matching module builds a feedback model based on the dynamic behavior and selection preferences of job seekers, and generates feedback information for updating the weight coefficient of hard information data. and the weight coefficient of soft information data , get the updated hard information data weight coefficient that meets the constraint conditions and updated soft information data weight coefficient ; Based on the updated hard information data weight coefficient and updated soft information data weight coefficient Calculate the updated job seeker suitability score and updated enterprise suitability scores , calculate the updated matching score , used to return to the matching model to obtain a new optimal matching result; The feedback information includes: the job seeker's score in the information dimension of the positions that were not successfully matched in the first round, the automatically assigned secondary positions, and the newly selected positions; the constraints are ; The weight coefficient of updating the hard information data and the weight coefficient of soft information data The methods include: ; In the formula, is the learning rate; k is the index of the information dimension; It represents the sum of the absolute values ​​of the difference values ​​of the information dimension scores in the hard information data; n is the total number of information dimensions in the hard information data; is the difference value of the information dimension, and the calculation formula is: ; In the formula, Score the information dimension of the secondary position, Score the dimensions of the new position; ; In the formula, represents the sum of the absolute values ​​of the difference values ​​of the information dimension scores in the soft information data; m is the total number of information dimensions in the soft information data; Scoring the match between job seekers and companies When performing calculations, define the KNN algorithm to optimize the matching score Optimization model of Define Euclidean distance as a measure between job seekers and companies, and calculate the mathematical distance between job seeker feature vectors and job feature vectors.

2. According to claim 1, a job resume intelligent matching system based on multi-dimensional analysis is characterized in that: The resume information data and recruitment information data both include hard information data and soft information data; wherein, the information dimensions of the hard information data include: educational information, work information, skill information and demand information; the information dimensions of the soft information data include: career interests, career direction, cultural fit and career stability.

3. According to claim 2, a job resume intelligent matching system based on multi-dimensional analysis is characterized in that: Hard Scoring of Job Applicants , Soft Rating of Job Applicants , Enterprise hard scoring and enterprise soft rating , the calculation formula is: ； In the formula, Score the job applicant's academic qualifications. Rating job seekers jobs, Score job applicant skills, Score job seekers' needs. is the influence coefficient of the corresponding information dimension; ； In the formula, Score job applicant interest, Score the job seeker's orientation, Score the culture of job seekers. Score the stability of job seekers. is the influence coefficient of the corresponding information dimension; ； In the formula, Score the company's academic qualifications. Rate the company's work. Score the skills of your business. Score business needs; ； In the formula, Score the interest of the business, Score the direction of the business, Rate your company culture. Score the stability of your business.

4. According to claim 3, a job resume intelligent matching system based on multi-dimensional analysis is characterized in that: The method of constructing the optimization model includes: constructing a feature vector according to the information dimension score, normalizing the information dimension score, mapping all feature values ​​to the [0,1] interval, and obtaining a hard information feature vector and soft information feature vector ; Using hard information feature vector and soft information feature vector Constructing a comprehensive feature vector ; Calculate the mathematical distance between the job applicant feature vector and the position feature vector. The calculation formula is: distance ; In the formula, represents the job applicant feature vector, ; represents the position feature vector, ; n is the number of features in the vector; , is the value of the i-th eigenvalue, belonging to the job seeker eigenvector and the position feature vector The i-th dimension of Calculate each sample and its target sample Distance , and record the distances in order And the corresponding sample index i; according to the calculated distance d value, all samples are sorted in ascending order to obtain a sorted list ; In the formula, is the sorted distance; is the minimum distance after sorting, the corresponding sample is the target sample The closest sample; is the maximum distance after sorting, the corresponding sample is the target sample The farthest sample; select the first K samples with the shortest distance from the sorted list List to construct the nearest neighbor set ; Assign weights to each nearest neighbor sample , the calculation formula is: ; In the formula, is the distance to the ith neighbor; Optimize the job seeker's suitability score And enterprise adaptability , the calculation formula is: ; In the formula, To optimize the suitability score of job seekers; represents the job seeker fitness score of the i-th neighbor; ; In the formula, The adaptability score for the optimized enterprise; represents the enterprise fitness score of the i-th neighbor; Using the optimized job seeker suitability score And the optimized enterprise adaptability score Calculate optimized match scores between job seekers and companies .

5. According to claim 4, a job resume intelligent matching system based on multi-dimensional analysis is characterized in that: The matching model is constructed in the following manner: Define the format of the matching matrix M: The rows of the matching matrix M represent each job applicant ; The columns of the matching matrix M represent each position ; Matching matrix M elements Representing job seekers With position Optimized match score of ; Based on the elements of each row in the matching matrix M size, determine each job seeker Priority list of positions ; Based on the elements of each column in the matching matrix M The size of each position Priority list of job seekers ; Set input as a priority list of positions and a priority list of job seekers ; Set the output to the optimal matching result; Initialize the matching state and perform step-by-step matching, repeating the step-by-step matching operation until all job seekers are matched. All matched or all positions are occupied, and all job seekers and Position Until there are no more better options; output the optimal matching result without any obstruction pairs.

6. The intelligent job resume matching system based on multi-dimensional analysis according to claim 5 is characterized in that: The data matching module will Input into the weight dynamic adjustment model and compare with the preset threshold T; if , then increase the weight of the information dimension with index k; if , the adjustment of the weight by the information dimension with index k is ignored.

7. A matching method according to any one of claims 1 to 6, characterized in that: The following steps are involved: Step 1: Collect the resume information data of job seekers and the recruitment information data of enterprises, and extract the information dimensions; Step 2: Calculate the applicant's suitability score and the company's suitability score based on the information dimension; Step 3: Calculate the matching score between job seekers and companies based on the job seeker suitability score and the company suitability score, compare the matching score with the preset threshold score, and give priority to matching job seekers and companies with matching scores greater than or equal to the threshold score.

Images