Personalized Job Recommender System Based On Behavior Analysis

With the rapid development and the widespread application of Internettechnology, online job seeking and recruiting have gradually become the mostimportant career choice approach. However, faced with a flood of career information, job seekers have to spend a lot of time to find useful information for a suitable job.The development of personalized recommendation technology provides a way tosolve the problem of information overload. The approach improves job websitesusability and the user experience.The paper studies the personalized job recommendation technology for collegestudents, which is based on the ongoing research project, a province talent websiterecruitment data and Web log from the website. The study follows the idea of "jobseekers with similar behavior should have similar job ". It puts forward the behaviorinfluence factor and the behavior analysis strategy to personalized job recommendersystem.Finally, an instance test shows that the effectiveness of this approach.The main research contents and work are listed below:1. This paper introduces the thought of behavior analysis into personalizedcase-based recommendation. According to analysis of the process behaviors of jobseekers searching job online, the behaviors of job seekers can be divided intoregistration behavior, history application behavior and browsing behavior. Based onthe case-based reasoning thought, that is,"job seekers with similar behavior shouldhave similar job ", the study introduces interest similarity in CBR, and presents thebehavior-intrest model of job seekers which based on the results of recruitment dataand Web log analysis. It overcomes the limitations in the former model of job seekerswhich only considers demographic properties and social competitiveness properties.2. The personalized job CBR recommendation uses the global similarity andlocal similarity to calculate the similarity between cases of job seekers. However, itdoes not solve the problem of weight in the previous work of the research project. Thepaper introduces the technology of structural equation model into CBR. It establishesa structural equation model which reflects the relationship between a set of job seeker’s factors and the success of job searching. Then the hidden variables amongjob seeker’s factors and their potential relationship is found through the use ofconfirmatory factor analysis. This thesis extracts local similarity weights from thestructural equation model which is used in the CBR similarity calculation. This studyshows that it improves the precision and quality of recommendation. It gives thetheoretical exploration for further study on other behavior.3. From the perspective of statistic and data mining, this paper summarizes thegeneral methods of factor analysis. Moreover, it classifies the case attributesaccording to data types of properties of the cases, and summarizes optional localsimilarity algorithms.4. In order to overcome the shortcomings in job seekers’ professional similaritycalculation with string comparison algorithm which was proposed in the past researchof the project, the paper pro-poses a new similar algorithm based on the similaritybetween taxonomic relation and non-taxonomic relation from the idea of ontologysimilarity. It solves the problem of similarity calculation of job seeker’s profession,and improves the accuracy of the similarity calculation.5. In the R environment and SQL Server2008database management system， Apackage of recommendation based on CBR is designed and implemented. It alsoperforms a test for the validity of the personalized job recommender system whichbased on the analysis of behaviors by a set of instances.