The labor market is a complex web of interactions between jobs and applicants. In this paper, the low -dimensional vector space will be constructed through a graph-convolution process, in which various jobs and applicants can be projected into this space. This embedding is then applied for solving various related tasks. Job recommendation is performed via link prediction, in which the graph-convolution embedding has a better performance than other alternatives models. Job similarity is performed via embedding proximity between jobs. Moreover, sector group is discovered via hierarchical clustering on the embedding distance. Finally, the labor market analysis is performed by comparing the supply and demand in each sector.
展开▼