PROJECT TITLE :

A Unified Framework for User Identification across Online and Offline Data

ABSTRACT:

Identification of users across multiple datasets has a wide variety of potential applications, which is one reason why there has been a growing body of research published on this subject over the past few years. However, the majority of previous research has focused on user identification using a single input data type. For example, (I) identifying a user across multiple social networks using online data, and (II) detecting a single user from heterogeneous trajectory datasets using offline data. Both of these approaches use online data. In contrast to other works, the one we are proposing in this paper is a framework for user identification that can be applied to both online and offline datasets. By creating a mapping from IP addresses to physical locations, we are able to establish connections between these two categories of data. We propose a novel framework that consists of three steps as a potential solution to this problem. To begin, we begin by mapping IP addresses into specific physical location distributions by employing a clustering method that is based on the locations of IP addresses. Second, in order to cut down on the amount of space required and the amount of time it takes to run the computation, we suggest a new pairwise index. In the third and final step, we use a method called learning-to-rank to combine the results of the previous two steps' collection of multiple features. We design experiments on the basis of our framework in order to demonstrate the effectiveness of our framework in terms of its efficiency (both in terms of time and space) as well as the precision and recall of our approach in comparison to other methods.