cube 2 net : E icient ality Network Construction with Data Cube Organization

Networks are widely used to model structured data and enable various downstream applications. However, in the real world, most data are structureless, and the assumption of a given network for each particular task is o‰en invalid. In this work, given a set of objects, we propose to leverage data cube to organize its enormous ambient data. Upon that, we further provide a reinforcement learning algorithm to automatically explore the cube structure and eciently select appropriate data for the construction of a quality network, which can facilitate various tasks on the given set of objects. With extensive experiments of two classic networkmining tasks on di‚erent real-world large datasets, we show that our proposed cube2net pipeline is general, and much more e‚ective and ecient in quality network construction, compared with other methods without the leverage of data cube or reinforcement learning.