Understanding Service Integration of Online Social Networks: A Data-Driven Study

TL;DR

This study analyzes how cross-site linking in online social networks like Medium influences social graph structure and user influence, and proposes a machine learning model to identify influential users for better service integration.

Contribution

It provides the first comprehensive analysis of cross-site linking effects on social graph structure and introduces a high-accuracy model to predict influential users from other OSNs.

Findings

Cross-site linking changes social graph structure and attracts new users.

Most new users do not become high PageRank users.

A machine learning model predicts high PageRank users with high accuracy.

Abstract

The cross-site linking function is widely adopted by online social networks (OSNs). This function allows a user to link her account on one OSN to her accounts on other OSNs. Thus, users are able to sign in with the linked accounts, share contents among these accounts and import friends from them. It leads to the service integration of different OSNs. This integration not only provides convenience for users to manage accounts of different OSNs, but also introduces usefulness to OSNs that adopt the cross-site linking function. In this paper, we investigate this usefulness based on users' data collected from a popular OSN called Medium. We conduct a thorough analysis on its social graph, and find that the service integration brought by the cross-site linking function is able to change Medium's social graph structure and attract a large number of new users. However, almost none of the new users would become high PageRank users (PageRank is used to measure a user's influence in an OSN). To solve this problem, we build a machine-learning-based model to predict high PageRank users in Medium based on their Twitter data only. This model achieves a high F1-score of 0.942 and a high area under the curve (AUC) of 0.986. Based on it, we design a system to assist new OSNs to identify and attract high PageRank users from other well-established OSNs through the cross-site linking function.