INCORPORATING USER ACTIVITY DATA FOR IMPROVED USER ENTITY RESOLUTION

Abstract

Widespread adoption of information systems has created digital data traces concerning the users, their relationships and personal activities. Interconnecting the digital data traces of a user from various information systems is known as User Entity Resolution (UER). Since prior research has loosely focused on incorporating user activity data for UER, the intention of this thesis is to develop methods that would enable the incorporation of user activity data in UER to further improve the accuracy. Prior researchers have conducted UER research in two directions, approximate matching/blocking and pairwise matching within the blocks. Acknowledging the limitations and differentiation power of user activity data, the thesis derives methods for using user activity data in these two directions. The thesis presents a novel latent representation based blocking method, a deep learning based cross domain transfer model and various extensions of the model to suit cold start scenario, auxiliary data and network relationships.