To improve the accuracy of job recommendations,a multi-behavioral job recommendation integrating knowledge graph (MB-JRIKG) was proposed based on the implicit preference information of both job seekers and recruiters in the process of [DK]"browsing job positions → submitting resumes → receiving feedback from recruiters". This method constructed a knowledge graph of job seeking based on real job data,and proposed a multi behavior preference propagation strategy based on preference propagation theory. [DK]"Recruiter recognition" in each stage of job seeking was set as the target behavior,and the auxiliary behaviors of job seekers browsing job positions and submitting resumes were set to comprehensively predict job seekers′ preferences. Firstly,the user′s historical records under different behaviors were used as user-interested seed sets,and preference propagation was carried out along the relationships between nodes in the knowledge graph to infer the user′s potential preferences and enhance user representation; Then,the user representation vector and position representation vector were input into the prediction function to calculate the interaction probability of the user under each behavior type,and the weighted sum was used as the interaction probability of the target behavior. Finally,the click-through rate prediction experiment was conducted using the competition dataset of Alibaba′s human job intelligent matching. The results show that compared with the four benchmark models (MF,XGBoost,KGCN and RippleNet),MB-JRIKG achieves a 0014 5 improvement in AUC and a 0028 8 improvement in ACC over the sub optimal benchmark model RippleNet,verifying the effectiveness of the model and achieving full utilization of data. This model effectively combines the interactive behavior of both parties seeking employment for recommendation,introduces attribute association of job knowledge graph,and has reference value for achieving personalized job recommendation.