A Data-Based Approach to Discovering Multi-Topic Influential Leaders

نویسندگان

  • Xing Tang
  • Qiguang Miao
  • Shangshang Yu
  • Yining Quan
چکیده

Recently, increasing numbers of users have adopted microblogging services as their main information source. However, most of them find themselves drowning in the millions of posts produced by other users every day. To cope with this, identifying a set of the most influential people is paramount. Moreover, finding a set of related influential users to expand the coverage of one particular topic is required in real world scenarios. Most of the existing algorithms in this area focus on topology-related methods such as PageRank. These methods mine link structures to find the expected influential rank of users. However, because they ignore the interaction data, these methods turn out to be less effective in social networks. In reality, a variety of topics exist within the information diffusing through the network. Because they have different interests, users play different roles in the diffusion of information related to different topics. As a result, distinguishing influential leaders according to different topics is also worthy of research. In this paper, we propose a multi-topic influence diffusion model (MTID) based on traces acquired from historic information. We decompose the influential scores of users into two parts: the direct influence determined by information propagation along the link structure and indirect influence that extends beyond the restrictions of direct follower relationships. To model the network from a multi-topical viewpoint, we introduce topic pools, each of which represents a particular topic information source. Then, we extract the topic distributions from the traces of tweets, determining the influence propagation probability and content generation probability. In the network, we adopt multiple ground nodes representing topic pools to connect every user through bidirectional links. Based on this multi-topical view of the network, we further introduce the topic-dependent rank (TD-Rank) algorithm to identify the multi-topic influential users. Our algorithm not only effectively overcomes the shortages of PageRank but also effectively produces a measure of topic-related rank. Extensive experiments on a Weibo dataset show that our model is both effective and robust.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

A Knowledge Management Approach to Discovering Influential Users in Social Media

A key step for success of marketer is to discover influential users who diffuse information and their followers have interest to this information and increase to diffuse information on social media. They can reduce the cost of advertising, increase sales and maximize diffusion of information.  A key problem is how to precisely identify the most influential users on social networks. In this pape...

متن کامل

Identifying Topical Opinion Leaders in Social Community Question Answering

Social community question answering (SCQA) sites not only provide regular question answering (QA) service but also form a social network where users can follow each other. Identifying topical opinion leaders who are both expert and influential in SCQA becomes a hot research topic. However, existing works focus on either using knowledge expertise to find experts for improving the quality of answ...

متن کامل

An Effective Tool for Recommending Opinion Leaders in Social Networking Service

In this paper, we propose and present a practical tool for social networking services that can be used to help users find the opinion leaders or influential users in a certain community. This tool can act as a front-end to existing social networking services such as Twitter and gives a list of influential users in their network given a topic. Our approach differs from most of existing recommend...

متن کامل

A review of text mining approaches and their function in discovering and extracting a topic

Background and aim: Four text mining methods are examined and focused on understanding and identifying their properties and limitations in subject discovery. Methodology: The study is an analytical review of the literature of text mining and topic modeling.  Findings: LSA could be used to classify specific and unique topics in documents that address only a single topic. The other three text min...

متن کامل

ACQR: A Novel Framework to Identify and Predict Influential Users in Micro-Blogging

As key roles of online social networks, influential users in micro-blogging have the ability to influence the attitudes or behaviour of others. When it comes to marketing, the users’ influence should be associated with a certain topic or field on which people have different levels of preference and expertise. In order to identify and predict influential users in a specific topic more effectivel...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره 11  شماره 

صفحات  -

تاریخ انتشار 2016