CN-121998298-A - Bidirectional employment recommendation method based on professional development trend modeling

CN121998298ACN 121998298 ACN121998298 ACN 121998298ACN-121998298-A

Abstract

The invention discloses a bidirectional employment recommendation method based on professional development trend modeling, which belongs to the field of employment recommendation methods and comprises the steps of multi-source data acquisition and preprocessing, double-portrait construction, professional development trend modeling, bidirectional matching and recommendation. The invention converts the hidden requirement into the quantifiable feature vector by constructing the job seeker portrait containing the family features and the post portrait organizing the ecological features, thereby solving the defect of single dimension of the portrait in the prior art. Experiments prove that after the ecological characteristics of the household tissues are introduced, the matching rate of the matching result and the actual employment adaptation degree is improved by more than 35%, and the problems of 'surface layer matching but actual inappropriateness' are effectively reduced.

Inventors

Xiong Quanlang
ZHOU JUNJIE
GAO ZEJIN
YANG QIONG
HUANG WUCHENG
LIU JINGJING

Assignees

武汉梦软科技有限公司
武汉软件工程职业学院(武汉开放大学)

Dates

Publication Date: 20260508
Application Date: 20251225

Claims (10)

1. The bidirectional employment recommendation method based on professional development trend modeling is characterized by comprising the following steps of: S1, multi-source data acquisition and preprocessing, namely acquiring dimension data of job seekers, dimension data of recruiters and macroscopic trend data, and cleaning, removing redundancy and standardizing the data; S2, double-portrait construction, namely constructing a job seeker portrait comprising family characteristics based on the preprocessed job seeker dimension data, and constructing a post portrait comprising organization ecological characteristics based on the preprocessed recruiter dimension data; s3, modeling occupational development trend, namely fusing macroscopic trend data, constructing a trend model through time sequence feature extraction and bidirectional development path prediction, and outputting a job seeker growth prediction result and a post demand evolution prediction result; S4, two-way matching and recommendation, namely, constructing a two-way matching model, inputting prediction results of the job seeker portrait, the post portrait and the trend model, calculating two-way fitness through dynamic weight adaptation, and generating and outputting a job seeker recommendation list and an enterprise candidate list according to the ranking of the fitness.
2. The method according to claim 1, wherein in step S1, the job seeker dimension data includes family structure and responsibility data, economic capital data, culture and value view data, interpersonal interaction mode data, background and resource data and boundary fusion demand data, the recruiter dimension data includes post hard demand data, interpersonal ecological data, boundary management policy data and development ecological data, and the macroscopic trend data includes industry report data, policy file data and market salary fluctuation data.
3. The bidirectional employment recommendation method based on professional development trend modeling of claim 1 is characterized in that in the step S2, the construction process of job seeker portraits comprises the steps of carrying out 0-1 coding, Z-score standardization and Likett scale conversion on family feature data, carrying out dimension reduction processing through principal component analysis, and outputting 128-dimensional family feature vectors, wherein the construction process of job portraits comprises the steps of carrying out keyword weight extraction and industry benchmark normalization on organization ecological feature data, and outputting 128-dimensional organization feature vectors through differential Transformer noise reduction processing.
4. The bidirectional employment recommendation method based on occupational development trend modeling according to claim 1, wherein in the step S3, the time series feature extraction adopts a time series feature pyramid network of a DMU-Net architecture, long-term trends are deeply identified through capturing seasonal fluctuations of occupational demands, multi-scale features are fused by combining a cross-layer attention mechanism, and the bidirectional development path prediction comprises simulating a promotion probability and a capability growth curve of job seekers based on a Markov decision process of reinforcement learning, and modeling ecological prejudgement demand changes of the posts of an enterprise based on a graph neural network.
5. The bidirectional employment recommendation method based on occupational development trend modeling, which is disclosed in claim 1, is characterized in that in step S4, the bidirectional matching model comprises a double-view encoder and an adaptive feature fusion layer, the double-view encoder respectively processes job seekers images and post images, the job seekers view angle strengthens the matching degree calculation of posts and occupational targets, the enterprise view angle strengthens the matching degree calculation of candidates and post future demands, and the adaptive feature fusion layer fuses two core complaints based on an attention mechanism.
6. The bidirectional employment recommendation method based on professional development trend modeling of claim 1, wherein in the step S4, the specific process of dynamic weight adaptation is that a dynamic multi-key value memory matrix is introduced, matching weights are adjusted according to the professional development stage of job seekers and the enterprise industry fluctuation rate, post culture system weights are improved for job seekers in a professional exploration stage, professional transformation target matching degree is emphasized for job seekers in a stable stage, and candidate skill migration capability weights are improved for high fluctuation industries.
7. The bidirectional employment recommendation method based on professional development trend modeling according to claim 1, further comprising the step S5 of dynamically optimizing a model, wherein parameters of a trend model and a bidirectional matching model are adjusted in real time through online learning based on retention after job seekers enter, promotion conditions and enterprise satisfaction feedback data, and model parameters are updated through migration learning by combining the industry trend data in a quarter.
8. The bidirectional employment recommendation method based on professional development trend modeling according to claim 1, wherein in step S2, the family structure and responsibility data are quantified by a support burden coefficient and child support requirements, the economic capital data are quantified by a family annual income share and professional trial-and-error period, and the interpersonal ecological data are quantified by a team communication mode and a lead management style tag code.
9. The bidirectional employment recommendation method based on professional development trend modeling according to claim 1, wherein the graph neural network adopts an overflow-intersection model, takes home characteristics of job seekers as node attributes, predicts forward overflow effects of candidates on teams, and predicts 3-year retention and contribution values in combination with home-work gain matching degree.
10. The method for bidirectional employment recommendation based on occupational development trend modeling according to claim 1, wherein in the step S4, the calculation formula of the bidirectional fitness is bidirectional fitness=α× (job seeker feature vector, post feature vector) +β×growth adaptation component+γ×value adaptation component, where α, β, γ are dynamic weights, initial values α=0.4, β=0.3, γ=0.3, and real-time calibration is performed by bayesian optimization.

Description

Bidirectional employment recommendation method based on professional development trend modeling Technical Field The invention relates to the field of employment recommendation methods, in particular to a bidirectional employment recommendation method based on professional development trend modeling. Background With the digitized development of employment markets, employment recommendation methods have evolved from traditional artificial matching to intelligent algorithm recommendation. In the prior art, employment recommendation is mostly based on the fact that explicit characteristics such as skills and academia of job seekers are matched with the skill demands of posts, part of methods introduce machine learning technology, recommendation optimization is achieved through keyword extraction and similarity calculation, recommendation priority is determined through comparison of evaluation keyword duty ratio of a waiter and an employment, a matching database is built, bidirectional data matching analysis is achieved through a machine learning training model, matching efficiency is improved, meanwhile, part of technologies can be combined with industry trend data, short-term fluctuation of post demands is captured through methods such as time sequence analysis, and recommendation decision is assisted. The prior art has obvious limitations, the long-term adaptation requirements of job seekers and enterprises are difficult to meet, firstly, the construction dimension of the image is single, the image of the job seeker focuses on dominant skills and histories only, hidden factors such as family structures, economic capital, value and the like are not included, the job image is limited to hard skill requirements, organization environment features such as interpersonal ecology, boundary management policies and the like are absent, the image cannot reflect real adaptation trends of the two parties, secondly, the trend modeling capability is insufficient, the prior art focuses on short-term job demand fluctuation, a bidirectional prediction mechanism of the growth path of the job seeker and the ecological evolution of the job position of the enterprise is not established, the adaptation stability of 3-5 years of the dimension cannot be pre-determined, thirdly, the matching mechanism is unidirectional, the matching degree is calculated from a single view, the dynamic adjustment weight is not carried out according to the development stages of the two parties, the problem that the recommended result is 'skill matching but long-term adaptation is poor' is caused, the retention rate of the enterprise is lower than the satisfaction degree of the job position, fourthly, the model lacks dynamic optimization, the closed loop adjustment mechanism based on actual job entry feedback is not established, and the dynamic adaptation trend and the dynamic change of the two parties and the demands are difficult to adapt to the dynamic change of the industries. Disclosure of Invention In order to overcome the problems, the invention aims to provide a bidirectional employment recommendation method based on professional development trend modeling, and aims to solve the problems provided by the background technology. For this purpose, the invention adopts the following specific technical scheme: According to one aspect of the invention, a bidirectional employment recommendation method based on professional development trend modeling is provided, which comprises the following steps: S1, multi-source data acquisition and preprocessing, namely acquiring dimension data of job seekers, dimension data of recruiters and macroscopic trend data, and cleaning, removing redundancy and standardizing the data; S2, double-portrait construction, namely constructing a job seeker portrait comprising family characteristics based on the preprocessed job seeker dimension data, and constructing a post portrait comprising organization ecological characteristics based on the preprocessed recruiter dimension data; s3, modeling occupational development trend, namely fusing macroscopic trend data, constructing a trend model through time sequence feature extraction and bidirectional development path prediction, and outputting a job seeker growth prediction result and a post demand evolution prediction result; S4, two-way matching and recommendation, namely, constructing a two-way matching model, inputting prediction results of the job seeker portrait, the post portrait and the trend model, calculating two-way fitness through dynamic weight adaptation, and generating and outputting a job seeker recommendation list and an enterprise candidate list according to the ranking of the fitness. Optionally, in step S1, the job seeker dimension data includes family structure and responsibility data, economic capital data, cultural and value view data, interpersonal interaction mode data, background and resource data and boundary fusion demand data, the recruiter dimension da