Learning and Multi-Objective Optimization for Automatic Identity Linkage

There has been exponential growth on the Internet connectable and connected devices in recent years. With the world becoming more connected than ever, each individual user accesses services from a range of devices, including personal desktops and laptop computers, tablets, mobile devices, vehicles, and entertainment systems. One fundamental problem is to identify the user using the fragmentation of identity and consumer profiles across these connected devices. Adobe Identity Graph solves this fundamental problem for the digital marketing domain. An Identity graph is built by discovering identity relationships using both online data traffic and offline data logs to establish links between different identities allowing for richer insights into the consumer. In this paper, we explore using a machine learning based approach for Identity Graph. In particular, we focus on linking identities using structured learning and multi objective optimization. Our proposed solution can help in improving current approaches in to give a complete holistic view of the consumer. Additionally, it can help learning about the consumer over time. Towards automatic identity linkage of the same natural person, we can study to construct statistical learning method based on massive user access logs. Our evaluation results show that our proposed approach can be very effective in linking identities and achieving a practical solution on a real data set.