Construction Of Job Competency Model And Optimization Of Enterprise Recruitment Based On Text Mining

This thesis has successfully constructed a job competency model in the field of human resource management by using the relevant theories and techniques of text mining.The job competency model includes a series of job competency elements,which are divided into two parts: explicit and implicit.As the cornerstone and pillar of human resources work,it occupies an important position and plays an important role in the daily work of enterprises,especially the recruitment process.However,the construction cost of traditional competency model is high,the data collection cycle is long,and the evaluation results are subjective,which are not conducive to enterprises to carry out this work,so the main purpose of this paper is to build a job competency model with text mining as the main technical means.By crawling and cleaning the data of mainstream recruitment websites,the paper selects the corresponding texts of job responsibilities and job requirements as the main research objects,and takes the topic modeling and text clustering in unsupervised algorithms as the main research methods.The model elements are all clustered and extracted,so as to successfully build a job competency model represented by the position of product manager,which is the core subject part of the article.Then,on the basis of this competency model,the recruitment information text is scientifically generated,and the related algorithm is used to calculate the text similarity with the corresponding resume,so as to achieve the purpose of optimizing the resume screening work.The paper mainly uses the LDA topic modeling technology and the improved Siamese network SBERT model based on Bert fine-tuning in the extraction of job competency elements.The LDA topic model is good at dealing with long texts.On the basis of not dividing the entire recruitment information,it mainly extracts the dominant elements in the job competency model.The SBERT model first uses sentence embedding to construct high-dimensional sentence vectors on short-text recruitment texts that are divided into independent single sentences,and then uses umap algorithm to reduce dimensionality and HDBSCAN algorithm for clustering,and successfully dig the hidden elements in depth.Topic modeling and text clustering complement each other to complete the construction of the competency model.Finally,the paper uses the TF-IDF algorithm to analyze the cosine similarity,and proposes a method to calculate the similarity between resumes and recruitment texts based on the competency model,so as to optimize the related resume screening work.This thesis has successfully constructed the competency model of relevant positions from the perspective of statistical text mining,and the effect is good.It can meet the requirements of mainstream enterprises for employment and has good guiding significance for the human resources management of enterprises.The text similarity calculation based on this model can efficiently handle the selection and screening of talents,which is worthy of further research.