Employment information matching method and system based on data analysis

Abstract

The invention discloses an employment information matching method and system based on data analysis, and particularly relates to the field of employment matching, and the method comprises the steps: collecting structured and unstructured data, and carrying out the semantic feature extraction through employing a BERT model and BiLSTM-CRF; in the preprocessing stage, entity standardization is realized through a knowledge graph, and a job seeker portrait and post model including a skill matrix and an occupational development trajectory is constructed; in the feature engineering stage, extracting four core features of skill matching degree, salary expectation integrating degree, commuting tolerance and occupational development goodness of fit; the salary integrating degree is quantitatively evaluated through a bidirectional tolerance model, a random forest model with time decay is used for dynamic weight, feature weight is generated based on historical successful cases, and a real-time feedback mechanism is introduced to adjust a weight coefficient; finally, the matching degree function fuses the weighted features and the industry trend factors, and the model is continuously optimized through a three-level updating mechanism.

Description

Technical Field

[0001] This invention relates to the field of employment matching technology, and more specifically, to a method and system for matching employment information based on data analysis. Background Technology

[0002] In today's society, the job market presents a complex and dynamic situation. On the one hand, with the number of college graduates increasing year by year and the flow of talent between various sectors of society becoming more frequent, the scale of job seekers is constantly expanding, and their employment needs are becoming more diversified, covering multiple dimensions such as industry selection, job type, salary and benefits, work location, and career development prospects.

[0003] On the other hand, companies' demands for talent are constantly evolving. Against the backdrop of industrial upgrading and technological innovation, companies are demanding increasingly refined and specialized skills for their positions, with significant differences in recruitment standards across different industries and company sizes. However, current job information matching models struggle to meet these complex needs on both the supply and demand sides.

[0004] However, it still has some shortcomings in practical use. Traditional job information matching relies on simple keyword search. Job seekers filter information by entering keywords such as job title and industry on recruitment platforms. It lacks in-depth mining of multi-source heterogeneous data (such as industry trends, salary fluctuations, career development paths, etc.) and cannot respond to market changes in real time. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a data analysis-based employment information matching method and system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] S1: Obtain structured and unstructured employment data in real time from corporate recruitment databases, job seeker resume databases, and publicly available industry data platforms;

[0008] S2: Denoise the collected data, fill in missing values, recognize entities and vectorize the text to generate standardized job seeker profiles and corporate job models;

[0009] S3: Based on job requirements and job seeker abilities, extract four core features: skill matching degree, salary expectation matching degree, commuting tolerance degree, and career development alignment degree.

[0010] S4: Train a random forest model using historical successful matching data to generate dynamic weight coefficients for each feature;

[0011] S5: Push the matching results to the user terminal in descending order of score, and iteratively update the weight model based on user behavior data.

[0012] Preferably, in the multi-source data acquisition of S1, structured data (such as job title, salary range, education requirements, etc.) from the enterprise recruitment database is crawled in real time using a distributed crawler framework, while standardized fields (including years of work experience, skill certificates, educational background, etc.) from the job seeker resume database are also accessed. For the unstructured data portion, a multimodal data processing pipeline is adopted: First, the free text descriptions in the enterprise recruitment brochure (such as paragraphs of job responsibilities and job requirements) are semantically segmented, and long texts are decomposed into independently analyzable semantic units using a text segmentation algorithm based on an attention mechanism; Second, the project experience text in the job seeker resume is analyzed for contextual association, and named entity recognition technology is used to extract the key technology stack, business areas, and achievement indicators of the project, and establish a mapping relationship with the standardized skill tree; Finally, career evaluation data from platforms such as LinkedIn and Maimai are obtained through social network API interfaces, including non-standard content such as anonymous employee evaluation texts and industry discussion posts.

[0013] Preferably, in S2, the system employs a multi-layered cleaning and transformation process to standardize the original employment data. First, data denoising is performed. For structured data (such as numerical fields like salary range and years of work experience), outliers are eliminated through interquartile range detection and Z-score standardization. For text data, regular expressions are used to filter HTML tags, special characters, and meaningless stop words. Simultaneously, an attention-based text cleaning model identifies and removes duplicate job postings or resume content. In the missing value handling stage, a differentiated strategy is adopted based on data type: numerical fields (such as positions with undisclosed salaries) are estimated using the KNN nearest neighbor algorithm for the same industry and position; categorical fields (such as missing educational requirements) use a Bayesian network for probability inference based on other known fields; and textual missing values ​​(such as blank project descriptions) are marked as "unmarked" without filling to avoid introducing noise.

[0014] In the entity recognition stage, a hybrid NER model integrating domain knowledge was deployed: for enterprise job descriptions, a pre-trained legal-human resources joint model was used to identify welfare entities such as "five social insurances and one housing fund" and "equity incentives"; for resume text, a fine-grained recognizer based on BiLSTM-CRF was used to extract technical experience entities such as "3 years of SpringBoot experience" and "led projects with tens of millions of page views", and the ambiguity of easily confused skills such as "Java" and "JavaScript" was resolved by constructing an industry knowledge graph; all identified entities were mapped to a standardized skill ontology (e.g., "proficient in PS" was unified as "advanced Adobe Photoshop skills"), and structured labels with confidence scores were generated.

[0015] Preferably, in step S3, during the feature extraction stage, four core matching dimensions are constructed through quantitative analysis, and each dimension adopts a composite index calculation system.

[0016] In the skill matching degree assessment, a hierarchical evaluation model is adopted. First, the hard skill matching degree is calculated based on the knowledge graph. The specific calculation method is as follows:

[0017]

[0018] in, This is expressed as a hard skill match. Let the weight of the i-th skill in the job be represented as . This represents the job seeker's skill level. This represents the required proficiency level for the position. This indicates the job seeker's actual level;

[0019] The soft skills section is calculated based on the semantic similarity between the resume text and the job description. The specific calculation method is as follows:

[0020]

[0021] Where B represents the soft skills match, Represented as a text vector of the job description. Represented as a vector of resume text;

[0022] The resume text and job description text are transformed into high-dimensional vectors through natural language processing.

[0023] The final skill match is a weighted harmonic average of hard and soft skills, calculated as follows:

[0024]

[0025] Where C represents the final skill match, A represents the hard skill match, and B represents the soft skill match. This is represented as a hard skill weighting factor. ;

[0026] A two-way tolerance model is constructed, which considers the alignment between the company's budget range and the job seeker's expected salary range. The specific calculation method is as follows:

[0027]

[0028] Where D represents the salary expectation fit. This represents the lower limit of a job seeker's salary expectations. This represents the upper limit of the job seeker's salary expectation. This represents the lower limit of a company's budget, i.e., the minimum salary a company is willing to pay; This represents the company's budget ceiling, which is the highest salary the company is willing to pay. This represents the range of market salaries.

[0029] In commuting tolerance, the spatiotemporal cost calculation based on geographic information systems is specifically as follows:

[0030]

[0031] Where F represents commuting tolerance. This represents the actual commute time. This is expressed as the maximum tolerable commuting threshold. Expressed as urban traffic coefficient, Represented as remote work weight, This is expressed as the percentage of days spent working remotely.

[0032] Preferably, in step S4, during the dynamic weight allocation stage, a random forest ensemble learning model with time-sensitive correction is constructed based on historical successful matching data, and dynamic weight coefficients are generated through multi-dimensional feature importance analysis. Specifically, positive samples (including job matching records where job seekers were ultimately hired and retained for more than 3 months) are extracted from the historical successful matching case library, while a negative sample set (including cases with high matching scores but explicitly rejected by users or those who left within 30 days of hiring) is constructed, forming a time-stamped training dataset. , It is a four-dimensional feature vector, that is ; This indicates a successful match. Represented as the timestamp of the match occurrence;

[0033] The model employs time-decay weighted sampling, where the sampling weight of recent samples increases exponentially. The specific calculation method is as follows:

[0034]

[0035] in, This represents the sampling weight of a certain sample, and K represents the attenuation coefficient. This is represented by the timestamp corresponding to the d-th sample. Represented as the minimum time value among all samples, Represented as the maximum time value among all samples;

[0036] The attenuation coefficient K=2 ensures that the model pays more attention to recent changes in market preferences;

[0037] During the training of a random forest, when splitting each decision tree node, in addition to calculating the traditional Gini impurity, a feature timeliness factor is also introduced. The specific calculation method is as follows:

[0038]

[0039] Where L represents the adjusted impurity, and g represents the traditional Gini impurity. This is represented as an adjustment parameter. This represents the average effective period of the current feature in historical data. Represented as a time-sensitive factor;

[0040] By training 500 decision trees in parallel, the average importance score of each feature on the OOB data is output. After Softmax normalization, the basic weights are obtained. The specific calculation method is as follows:

[0041]

[0042] in, The basic weights of feature f are represented as follows. Represented as average importance score, This can be expressed as the sum of the exponential values ​​of all features. Represented as the original importance score for feature f. Take the natural index.

[0043] Preferably, in step S5, after the matching degree calculation is completed, the results are first processed in a scenario-based manner: for enterprise HR users, the simplified profiles of the top 50 candidates are displayed in descending order of MatchScore, and a "Job Competitiveness Analysis Panel" is provided to intuitively show the deviation of the job from the market average in terms of salary, skill requirements, etc.; for job seeker users, a progressive disclosure strategy is adopted, with the core information (salary, commute time, core skill matching degree) of the 3 jobs with the highest matching degree displayed on the first screen, and 20 recommended jobs arranged in descending order of MatchScore can be viewed after scrolling down, with a "Matching Advantage Radar Chart" to visualize the scores of each dimension;

[0044] To construct an effective feedback loop, a three-level weight update mechanism was designed: short-term updates are fine-tuned based on real-time user behavior; when the click-through rate of a certain job category deviates from the historical average by ±15% for four consecutive hours, it is immediately updated. Adjust the corresponding feature weights, where, This is expressed as the adjustment amount for the feature weights. Represented as the learning rate, This represents the click-through rate for a specific job category within the current time window (a continuous 4-hour period). This is represented as the historical average click-through rate; mid-term updates are performed weekly, using successful matching cases from the past 7 days (defined as records of companies sending interview invitations after viewing resumes and job seekers accepting) to retrain the random forest model and optimize the cross-feature weights of commuting tolerance and career development; long-term updates are performed monthly in conjunction with the macro report on the human resources market, adjusting the weight allocation of each sub-item in the calculation formula of the industry trend factor λ through expert rules.

[0045] The technical effects and advantages of this invention are as follows:

[0046] This invention utilizes distributed web crawlers to collect real-time enterprise recruitment data and job seeker resumes. It employs the BERT model and BiLSTM-CRF technology to process unstructured text, constructing standardized talent profiles and job models. The core innovation lies in designing a four-dimensional feature system: skill matching, salary expectation alignment, commuting tolerance, and career development alignment. Feature weights are dynamically adjusted using a random forest model, and an industry trend factor λ is introduced for real-time calibration. Matching results are pushed using a tiered strategy, incorporating user behavior data to establish a three-level feedback mechanism: real-time fine-tuning, mid-term optimization, and long-term adjustment.

[0047] In terms of matching accuracy, a multi-level cleaning and transformation process is adopted to generate standardized profiles and models, extract four types of core features and quantify them, and combine them with a random forest model to generate dynamic weights, which greatly improves the accuracy of matching job positions with job seekers and reduces information mismatch.

[0048] The model is highly adaptable, incorporating time decay weighted sampling, real-time feedback adjustment, and a three-level weight update mechanism, enabling the model to respond promptly to market changes and user behavior trends, dynamically optimize weights, and enhance its adaptability to the dynamics of the job market. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0050] Figure 2 This is a schematic diagram of the module connection of the present invention.

[0051] Figure 3 This is a schematic diagram illustrating the logic for extracting the four core features of this invention.

[0052] Figure 4 This is a schematic diagram of the user-side recommendation logic of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Please see Figure 1 As shown, this invention provides a data analysis-based employment information matching method, the method being as follows:

[0055] S1: Obtain structured and unstructured employment data in real time from corporate recruitment databases, job seeker resume databases, and publicly available industry data platforms;

[0056] In S1's multi-source data acquisition, a distributed crawler framework is used to capture structured data (such as job title, salary range, and education requirements) from enterprise recruitment databases in real time, while simultaneously accessing standardized fields (including years of work experience, skill certificates, and educational background) from job seekers' resume databases. For unstructured data, a multimodal data processing pipeline is employed: First, semantic segmentation is performed on the free text descriptions in enterprise recruitment brochures (such as job responsibilities and requirements paragraphs), using an attention-based text segmentation algorithm to decompose long texts into independently analyzable semantic units. Second, contextual analysis is performed on project experience text in job seekers' resumes, using named entity recognition technology to extract key technology stacks, business areas, and achievement indicators from projects, and establishing a mapping relationship with standardized skill trees. Finally, career evaluation data from platforms such as LinkedIn and Maimai is obtained through social network API interfaces, including anonymous employee evaluation texts, industry discussion posts, and other non-standard content. To address these heterogeneous and unstructured data, the system deployed a semantic feature extraction layer based on the BERT model. Specifically, this included: performing domain-adaptive pre-training on the recruitment brochure text; fine-tuning the BERT token embedding layer on an HR professional corpus to improve the representation accuracy of professional terminology; employing a BiLSTM-CRF joint model to extract fine-grained entity relations from resume project experience, identifying "technology-project-achievement" triples; constructing a sentiment analysis subunit for social media evaluations, outputting sentiment polarity scores and topic clustering features using BERT's [CLS] tag; and converting all extracted semantic features into 768-dimensional dense vectors and storing them in an Elasticsearch vector database, while retaining pointer indices from the original text for interpretability analysis. The data acquisition module used a Kafka message queue for real-time stream processing, ensuring that new recruitment information completed the entire process from data acquisition to feature extraction within 15 minutes.

[0057] S2: Denoise the collected data, fill in missing values, recognize entities and vectorize the text to generate standardized job seeker profiles and corporate job models;

[0058] In S2, the system employs a multi-layered cleaning and transformation process to standardize the raw employment data. First, it performs data denoising. For structured data (such as numerical fields like salary range and years of work experience), outliers are eliminated through interquartile range detection and Z-score standardization. For text data, regular expressions are used to filter HTML tags, special characters, and meaningless stop words. Simultaneously, an attention-based text cleaning model identifies and removes duplicate job postings or resume content. In the missing value handling stage, a differentiated strategy is adopted based on data type: numerical fields (such as positions with undisclosed salaries) are estimated using the KNN nearest neighbor algorithm for the same industry and position; categorical fields (such as missing educational requirements) use a Bayesian network for probability inference based on other known fields; and textual missing values ​​(such as blank project descriptions) are marked as "unmarked" without being filled to avoid introducing noise.

[0059] In the entity recognition stage, a hybrid NER model integrating domain knowledge was deployed: for enterprise job descriptions, a pre-trained legal-human resources joint model was used to identify welfare entities such as "five social insurances and one housing fund" and "equity incentives"; for resume text, a fine-grained recognizer based on BiLSTM-CRF was used to extract technical experience entities such as "3 years of SpringBoot experience" and "led projects with tens of millions of page views", and the ambiguity of easily confused skills such as "Java" and "JavaScript" was resolved by constructing an industry knowledge graph; all identified entities were mapped to a standardized skill ontology (e.g., "proficient in PS" was unified as "advanced Adobe Photoshop skills"), and structured labels with confidence scores were generated.

[0060] In the text vectorization stage, a hierarchical feature representation strategy is implemented. For text fragments dense with technical terms (such as "familiar with distributed system architecture design" in job requirements), the SBERT model is used to generate domain-optimized sentence vectors. For descriptive content (such as introductions to corporate culture), Doc2Vec is used to preserve paragraph-level semantic features. In particular, for key information such as salary and benefits, a dedicated numerical perception embedding model is trained (e.g., encoding "15-20k" into a hybrid vector containing quantile features). The final standardized job seeker profile includes: basic attribute vectors (education / experience, etc.), a skills matrix (technical stack mastery score), and a career development vector (job / industry career path). The corporate job model includes: a core requirement vector (hard requirements), a preference auxiliary vector (bonus points), and a competitiveness indicator (quantified salary / benefits).

[0061] All preprocessed data is stored in a graph structure in the Neo4j database. Nodes contain entities and their relation weights, while edge attributes store co-occurrence frequencies and temporal features, providing an interpretable feature base for subsequent matching calculations. The entire preprocessing process is scheduled using Airflow and includes a quality monitoring unit that automatically triggers reprocessing when data consistency falls below 95%.

[0062] S3: Based on job requirements and job seeker abilities, extract four core features: skill matching degree, salary expectation matching degree, commuting tolerance degree, and career development alignment degree.

[0063] In S3, during the feature extraction stage, four core matching dimensions are constructed through quantitative analysis, and each dimension adopts a composite index calculation system.

[0064] In the skill matching degree assessment, a hierarchical evaluation model is adopted. First, the hard skill matching degree is calculated based on the knowledge graph. The specific calculation method is as follows:

[0065]

[0066] in, This is expressed as a hard skill match. Let the weight of the i-th skill in the job be represented as . This represents the job seeker's skill level. This represents the required proficiency level for the position. This indicates the job seeker's actual level;

[0067] The soft skills section is calculated based on the semantic similarity between the resume text and the job description. The specific calculation method is as follows:

[0068]

[0069] Where B represents the soft skills match, Represented as a text vector of the job description. Represented as a vector of resume text;

[0070] The resume text and job description text are transformed into high-dimensional vectors through natural language processing.

[0071] The final skill match is a weighted harmonic average of hard and soft skills, calculated as follows:

[0072]

[0073] Where C represents the final skill match, A represents the hard skill match, and B represents the soft skill match. This is represented as a hard skill weighting factor. ;

[0074] A two-way tolerance model is constructed, which considers the alignment between the company's budget range and the job seeker's expected salary range. The specific calculation method is as follows:

[0075]

[0076] Where D represents the salary expectation fit. This represents the lower limit of a job seeker's salary expectations. This represents the upper limit of the job seeker's salary expectation. This represents the lower limit of a company's budget, i.e., the minimum salary a company is willing to pay; This represents the company's budget ceiling, which is the highest salary the company is willing to pay. This represents the range of market salaries.

[0077] For example, suppose a major internet company is hiring an AI architect, and the company's budget... , Salary expectations of job seekers , Market benchmark ;

[0078] Feasibility check: If the conditions are met, the main calculation process will begin.

[0079] Calculate the numerator:

[0080]

[0081] Calculate salary expectation fit: If the result is a perfect match, the output will be displayed.

[0082] Another example is corporate budgets. , Salary expectations of job seekers , Market benchmark ;

[0083] Feasibility check: If the conditions are met, the main calculation process will begin.

[0084] Calculate the numerator: 14

[0085]

[0086] Calculate salary expectation fit: If the calculated salary expectation match score is extremely low, then there is almost no room for negotiation.

[0087] The market salary range spans the width of the median salary range for the same position in the same region, for example, the difference between the 25th percentile and the 75th percentile;

[0088] In commuting tolerance, the spatiotemporal cost calculation based on geographic information systems is specifically as follows:

[0089]

[0090] Where F represents commuting tolerance. This represents the actual commute time. This is expressed as the maximum tolerable commuting threshold. Expressed as urban traffic coefficient, Represented as remote work weight, This is expressed as the percentage of days worked remotely;

[0091] The urban traffic coefficient is set at 0.8 for first-tier cities, 0.5 for second- and third-tier cities, and 0.3 for fourth- and fifth-tier cities.

[0092] Career development fit is calculated using a three-dimensional evaluation model. The specific calculation method is as follows:

[0093]

[0094] Where G represents the degree of alignment between career development and practical application. Represented as path growth value, Represented as capability adaptation value, Represented as industry inertia value;

[0095] The path growth value is calculated using the Pearson correlation coefficient between the job seeker's historical job promotion speed and the promotion path to the target position.

[0096] The competency fit score is calculated as the ratio of the slope of the job seeker's learning curve to the rate of technological iteration for the job.

[0097] The specific method for calculating industry inertia value is as follows:

[0098]

[0099] Where I represents the industry inertia value, This represents the cumulative years of experience in the target industry. This is indicated as a job requirement. This represents the distance between the job posting and the candidate's industry sub-sector. This is represented as the maximum allowed subdivision span threshold;

[0100] The default value is a fixed constant of 1, which is the maximum cross-domain range that a company can accept. Exceeding this value is considered as a mismatch between professional expertise and job requirements. This involves quantifying the distance between the two entities within their respective sub-sectors, such as "Internet finance". "Traditional banks" Therefore, the distance is |0.8−0.2|=0.6;

[0101] All features are normalized to the [0,1] interval using the sigmoid function, and the feature distribution shift is monitored in real time.

[0102] S4: Train a random forest model using historical successful matching data to generate dynamic weight coefficients for each feature;

[0103] In step S4, during the dynamic weight allocation phase, a random forest ensemble learning model with time-sensitive correction is constructed based on historical successful matching data. Dynamic weight coefficients are generated through multi-dimensional feature importance analysis. The specific process is as follows: First, positive samples (including job matching records where job seekers were ultimately hired and retained for more than 3 months) are extracted from the historical successful matching case library. At the same time, a negative sample set (including cases with high matching degree but explicitly rejected by users or those who left within 30 days of being hired) is constructed to form a time-stamped training dataset. , It is a four-dimensional feature vector, that is ; This indicates a successful match. Represented as the timestamp of the match occurrence;

[0104] The model employs time-decay weighted sampling, where the sampling weight of recent samples increases exponentially. The specific calculation method is as follows:

[0105]

[0106] in, This represents the sampling weight of a certain sample, and K represents the attenuation coefficient. This is represented by the timestamp corresponding to the d-th sample. Represented as the minimum time value among all samples, Represented as the maximum time value among all samples;

[0107] The attenuation coefficient K=2 ensures that the model pays more attention to recent changes in market preferences;

[0108] During the training of a random forest, when splitting each decision tree node, in addition to calculating the traditional Gini impurity, a feature timeliness factor is also introduced. The specific calculation method is as follows:

[0109]

[0110] Where L represents the adjusted impurity, and g represents the traditional Gini impurity. This is represented as an adjustment parameter. This represents the average effective period of the current feature in historical data. Represented as a time-sensitive factor;

[0111] By training 500 decision trees in parallel, the average importance score of each feature on the OOB data is output. After Softmax normalization, the basic weights are obtained. The specific calculation method is as follows:

[0112]

[0113] in, The basic weights of feature f are represented as follows. Represented as average importance score, This can be expressed as the sum of the exponential values ​​of all features. Represented as the original importance score for feature f. Take the natural index;

[0114] In this context, 0 is not used for exponentiation; it is a label and does not participate in the calculation.

[0115] To further enhance model adaptability, a real-time feedback adjustment mechanism is introduced: when a significant change trend is detected in user behavior of a certain feature in recent (30 days) (e.g., the click pass rate of commuting tolerance drops by more than 15%), local retraining is automatically triggered. New data within the sliding time window is used to incrementally update the subtrees of the affected features, generating adjusted dynamic weights. The specific calculation method is as follows:

[0116]

[0117] in, Represented as in time The final weight of the time feature f, The basic weights of feature f are represented as follows. This is expressed as an adjustment coefficient. This is expressed as the slope of the change of feature f within the time window;

[0118] middle, This is not an exponential operation. These are labels and do not participate in calculations.

[0119] The final matching score function is constructed based on weighted features. The specific method for calculating the final matching score is as follows:

[0120]

[0121] Where Q represents the final matching score. This is represented by the value of feature f. Represented as in time The final weight of the time feature f, This is expressed as an industry adjustment coefficient. This represents the industry trend adjustment item. If the demand for talent in a certain industry has increased recently, the value of q is positive; otherwise, the value of q is negative.

[0122] Feature f includes skill matching, salary expectation matching, commuting tolerance, and career development alignment.

[0123] S5: Push the matching results to the user terminal in descending order of score, and iteratively update the weight model based on user behavior data;

[0124] In S5, after the matching degree is calculated, the results are first processed in a scenario-based manner: for enterprise HR users, the simplified profiles of the top 50 candidates are displayed in descending order of MatchScore, and a "Job Competitiveness Analysis Panel" is provided to intuitively show the deviation of the job from the market average in terms of salary, skill requirements, etc.; for job seekers, a progressive disclosure strategy is adopted, with the core information (salary, commute time, core skill matching degree) of the 3 jobs with the highest matching degree displayed on the first screen, and 20 recommended jobs arranged in descending order of MatchScore can be viewed after scrolling down, with a "Matching Advantage Radar Chart" to visualize the scores of each dimension;

[0125] To construct an effective feedback loop, a three-level weight update mechanism was designed: short-term updates are fine-tuned based on real-time user behavior; when the click-through rate of a certain job category deviates from the historical average by ±15% for four consecutive hours, it is immediately updated. Adjust the corresponding feature weights, where, This is expressed as the adjustment amount for the feature weights. Represented as the learning rate, This represents the click-through rate for a specific job category within the current time window (a continuous 4-hour period). This is represented as the historical average click-through rate; mid-term updates are performed weekly, using successful matching cases from the past 7 days (defined as records of companies sending interview invitations after viewing resumes and job seekers accepting) to retrain the random forest model and optimize the cross-feature weights of commuting tolerance and career development; long-term updates are performed monthly in conjunction with the macro report on the human resources market, adjusting the weight allocation of each sub-item in the calculation formula of the industry trend factor λ through expert rules.

[0126] All updates were verified through A / B testing. The new weighting model was first rolled out to 5% of traffic in a phased rollout, and then fully rolled out once the conversion rate improvement was confirmed to be statistically significant.

[0127] Please see Figure 2As shown in this embodiment, it should be specifically noted that the present invention provides an employment information matching system based on data analysis, including the following modules:

[0128] Multi-source data acquisition module: Real-time acquisition of structured and unstructured employment data from enterprise recruitment databases, job seeker resume databases, and publicly available industry data platforms;

[0129] Data preprocessing module: Denoises, fills in missing values, identifies entities, and vectorizes text in the collected data to generate standardized job seeker profiles and corporate job models;

[0130] Multidimensional feature extraction module: Based on job requirements and job seeker abilities, extract four core features: skill matching degree, salary expectation matching degree, commuting tolerance degree, and career development alignment degree.

[0131] Dynamic weight allocation module: Trains a random forest model using historical successful matching data to generate dynamic weight coefficients for each feature;

[0132] Intelligent recommendation and feedback optimization module: Pushes matching results to the user terminal in descending order of score, and iteratively updates the weight model based on user behavior data.

[0133] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data analysis-based employment information matching method, characterized in that, include: S1: Obtain structured and unstructured employment data in real time from corporate recruitment databases, job seeker resume databases, and publicly available industry data platforms; S2: Denoise the collected data, fill in missing values, recognize entities and vectorize the text to generate standardized job seeker profiles and corporate job models; S3: Based on job requirements and job seeker abilities, extract four core features: skill matching degree, salary expectation matching degree, commuting tolerance degree, and career development alignment degree. S4: Train a random forest model using historical successful matching data to generate dynamic weight coefficients for each feature; S5: Push the matching results to the user terminal in descending order of score, and iteratively update the weight model based on user behavior data.

2. The employment information matching method based on data analysis according to claim 1, characterized in that: In the multi-source data acquisition of S1, structured data from the enterprise recruitment database is crawled in real time using a distributed crawler framework, while standardized fields from the job seeker resume database are also accessed. For the unstructured data portion, a multimodal data processing pipeline is adopted: First, the free text descriptions in the enterprise recruitment brochures are semantically segmented, and long texts are decomposed into independently analyzable semantic units using an attention-based text segmentation algorithm; Second, the project experience text in the job seeker resumes is analyzed for contextual association, and named entity recognition technology is used to extract the key technology stack, business domains, and achievement indicators in the projects, and a mapping relationship with the standardized skill tree is established.

3. The employment information matching method based on data analysis according to claim 1, characterized in that: In S2, a multi-layered cleaning and transformation process is used to standardize the original employment data. First, data denoising is performed. For structured data, outliers are eliminated through interquartile range detection and Z-score standardization. For text data, regular expressions are used to filter HTML tags, special characters, and meaningless stop words. Attention-based text cleaning models are used to identify and remove duplicate job postings or resume content. In the missing value handling stage, a differentiated strategy is adopted according to the data type: numerical fields are estimated using the KNN nearest neighbor algorithm for the same industry and job; categorical fields use Bayesian networks to make probability inferences based on other known fields; and missing text values ​​are marked as "unmarked" without being filled to avoid introducing noise.

4. The employment information matching method based on data analysis according to claim 1, characterized in that: In S3, during the feature extraction stage, four core matching dimensions are constructed through quantitative analysis, and each dimension adopts a composite index calculation system. In the skill matching degree assessment, a hierarchical evaluation model is adopted. First, the hard skill matching degree is calculated based on the knowledge graph. The specific calculation method is as follows: in, This is expressed as a hard skill match. Let the weight of the i-th skill in the job be represented as . This represents the job seeker's skill level. This represents the required proficiency level for the position. This indicates the job seeker's actual level; The soft skills section is calculated based on the semantic similarity between the resume text and the job description. The specific calculation method is as follows: Where B represents the soft skills match, Represented as a text vector of the job description. It is represented as a vector of resume text.

5. The employment information matching method based on data analysis according to claim 1, characterized in that: A two-way tolerance model is constructed, which considers the alignment between the company's budget range and the job seeker's expected salary range. The specific calculation method is as follows: Where D represents the salary expectation fit. This represents the lower limit of a job seeker's salary expectations. This represents the upper limit of the job seeker's salary expectation. This represents the lower limit of a company's budget, i.e., the minimum salary a company is willing to pay; This represents the company's budget ceiling, which is the highest salary the company is willing to pay. This represents the range of market salaries. In commuting tolerance, the spatiotemporal cost calculation based on geographic information systems is specifically as follows: Where F represents commuting tolerance. This represents the actual commute time. This is expressed as the maximum tolerable commuting threshold. Expressed as urban traffic coefficient, Represented as remote work weight, This is expressed as the percentage of days spent working remotely.

6. The employment information matching method based on data analysis according to claim 1, characterized in that: In step S4, during the dynamic weight allocation phase, a random forest ensemble learning model with time-sensitive correction is constructed based on historical successful matching data. Dynamic weight coefficients are generated through multi-dimensional feature importance analysis. Specifically, positive samples are extracted from the historical successful matching case library, while a negative sample set is constructed to form a time-stamped training dataset. , It is a four-dimensional feature vector, that is ; This indicates a successful match. Represented as the timestamp of the match occurrence; The model employs time-decay weighted sampling, where the sampling weight of recent samples increases exponentially. The specific calculation method is as follows: in, This represents the sampling weight of a certain sample, and K represents the attenuation coefficient. This is represented by the timestamp corresponding to the d-th sample. Represented as the minimum time value among all samples, This represents the maximum time value among all samples.

7. The employment information matching method based on data analysis according to claim 6, characterized in that: During the training of the random forest, the traditional Gini impurity is calculated when each decision tree node splits, and a feature timeliness factor is introduced. Through parallel training of 500 decision trees, the average importance score of each feature on the OOB data is output, and the basic weights are obtained after Softmax normalization. The specific calculation method is as follows: in, The basic weights of feature f are represented as follows. Represented as average importance score, This can be expressed as the sum of the exponential values ​​of all features. Represented as the original importance score for feature f. Take the natural index; When a significant trend in recent user behavior is detected for a certain feature, local retraining is automatically triggered. New data within the sliding time window is used to incrementally update the subtrees of the affected features, generating adjusted dynamic weights. The specific calculation method is as follows: in, Represented as in time The final weight of the time feature f, The basic weights of feature f are represented as follows. This is expressed as an adjustment coefficient. It is represented as the slope of the change of feature f within the time window.

8. The employment information matching method based on data analysis according to claim 1, characterized in that: In S5, after the matching degree is calculated, the results are first processed in a scenario-based manner: for enterprise HR users, the simplified profiles of the top 50 candidates are displayed in descending order of MatchScore, and a "Job Competitiveness Analysis Panel" is provided to show the deviation of the job from the market average in terms of salary and skill requirements; for job seekers, a progressive disclosure strategy is adopted, with the core information of the 3 jobs with the highest matching degree displayed on the first screen, and 20 recommended jobs arranged in descending order of MatchScore can be viewed after scrolling down, with a "Matching Advantage Radar Chart" to visualize the scores of each dimension.

9. A data analysis-based employment information matching system, using the data analysis-based employment information matching method as described in any one of claims 1-8, characterized in that: Multi-source data acquisition module: Real-time acquisition of structured and unstructured employment data from enterprise recruitment databases, job seeker resume databases, and publicly available industry data platforms; Data preprocessing module: Denoises, fills in missing values, identifies entities, and vectorizes text in the collected data to generate standardized job seeker profiles and corporate job models; Multidimensional feature extraction module: Based on job requirements and job seeker abilities, extract four core features: skill matching degree, salary expectation matching degree, commuting tolerance degree, and career development alignment degree. Dynamic weight allocation module: Trains a random forest model using historical successful matching data to generate dynamic weight coefficients for each feature; Intelligent recommendation and feedback optimization module: Pushes matching results to the user terminal in descending order of score, and iteratively updates the weight model based on user behavior data.

Images