DescriptionThis dissertation aims at developing effective and efficient data mining techniques to solve varied talent recruitment issues, reforming the overall process with respect to talent sourcing, screening, matching, and assessment. Intelligent talent recruitment has gained increasing attention due to the critical talent competitions and intensive talent mobilities over the years. Previous studies mainly focus on discovering conceptual and theoretical topics, while applications for supporting organizational decision making are still under-explored.
To this end, we propose several approaches purposed to not only help the people to make intelligent talent-related decisions but also obtain domain understandings through a multifaceted data-driven perspective. In particular, we first present a hierarchical career-path-aware neural network to study individuals’ job mobilities. In this work, two problems are predicted all together on the basis of one’s historical career paths: 1) who will the individual’s next employer? 2) How long will the individual stay with his/her next employer? Several job mobility patterns regarding working duration, firm types, and etc. are discovered simultaneously. Also, we propose an intelligent matrix factorization based framework to address job salary benchmarking tasks. In this work, we consider multiple contextual factors to improve the prediction accuracy, such as job responsibility, company features, work location, and the time the job wanted. Furthermore, we put forward a Non-parametric Dirichlet Process-based graphical model to address the “cold-start” problem for salary benchmarking, which also has superior interpretability associated with job responsibility and company.