面向简历文本的端到端岗位推荐算法研究

Research on the end-to-end job recommendation algorithm for resume texts

为增大应聘者进入招聘初试的概率,基于云南省某大型国有企业线下真实招聘数据构建数据集,对岗位推荐算法进行了实证研究。利用构建好的数据集,对研究岗位推荐算法进行研究,分别对机器学习算法中的随机森林、xgboost模型、GBDT模型、LightGBM 4种机器学习模型,以及深度学习中的卷积神经网络模型和BERT模型进行实验。对比6种模型的岗位推荐结果,BERT模型在岗位推荐过程中的表现最佳,推荐准确率可达88.12%,说明BERT模型可用于岗位推荐类数据集并可以取得较好的推荐效果。另外,BERT模型对输入数据的处理相对更少,是一种端到端的学习模型,可以更方便的应用于岗位推荐。

To increase the probability of candidates entering the initial recruitment interview.In this study,a dataset was built based on real on-site recruitment data of a large state-owned enterprise in Yunnan Province,and empirical research was carried out on job recommendation algorithms.Four kinds of machine learning models including random forest,xgboost model,GBDT model and LightGBM,as well as convolutional neural network model and BERT model in deep learning,were tested respectively by using the built dataset.Comparing the job recommendation results of six models,the BERT model performs the best in the job recommendation process,with a recommendation accuracy of 88.12%.This indicates that the BERT model can be used for job recommendation datasets and can achieve good recommendation results.In addition,the BERT model has relatively less processing of input data and is an end-to-end learning model that can be more conveniently applied to job recommendations.