Visual Analysis And Personalized Recommendation Of Job Information Based On Data Mining

The development of the Internet and terminal technology has realized data sharing,making online job search a mainstream way to find a job.But at the same time,there is also the problem of information overload.At present,most job recommendation systems do not take implicit feedback into consideration,and do not fully exploit the textual value of job descriptions.Therefore,this thesis proposes a visual analysis and personalized recommendation of job information based on data mining.The main work includes:（1）Position data is preprocessed and carried out visual analysis.Feature cleaning is performed on the real job data collected based on web crawler technology,and then visual analysis is carried out from multiple perspectives such as numerical distribution,single-factor analysis,multi-factor analysis and word cloud graph to generate a more intuitive job portrait,which lays the foundation for the feature processing stage in the subsequent modeling.（2）A multivariate dynamic personalized job recommendation model based on attention mechanism is constructed.By introducing an attention mechanism,combined with explicit feedback and implicit feedback from users,a dynamic talent portrait is constructed,and the weight is dynamically adjusted according to the degree of attention of job seekers to each feature.The feedforward neural network is then trained with the class reorganization method and the Keras framework,and the hyperparameter settings are explored by grid search method.（3）A personalized job recommendation model based on text topics is constructed.The set of users who are most similar to the target job seeker is calculated from the users who have submitted their resumes,and the TF-IDF algorithm is used to construct the job topic based on the text descriptions of the jobs posted by similar users.Finally,based on the normalized Google distance,the similarity between the job topic and the candidate job is calculated to obtain the final Top-N recommendation list.（4）The personalized recommendation system is implemented based on the Flask framework.The demand analysis and system design of the personalized recommendation system for job information are carried out,combined with the multi-dynamic personalized job recommendation model based on the attention mechanism and the personalized job recommendation model based on text topics constructed in this thesis.Finally,the system is implemented through Hadoop environment and Flask framework.The evaluation of the model shows that when the length of the Top-N recommendation list is 10,compared with the User Based Pearson Correlation Coefficient（UPCC）model,the Hit Ratio（HR）index and the Normalized Discounted Cumulative Gain（NDCG）index of the new model are significantly improved.The method in this thesis deepens the understanding of job characteristics,improves the success rate of recommendation,and then better meets the individual needs of job seekers.