The Propagation Background in Social Networks: Simulating and Modeling

  • Share:

Online social networks play a significant role in our daily lives, especially as a crucial type of information propagation channel. Obviously, understanding the spreading of topics, ideas, and memes in social networks could be helpful to cognize human behaviors and get commercial interests. Thus, large efforts have been made on the propagation related problems.

 

In general, the information propagation process reflects users′ decisions on retweeting/sharing the content items they received, which could be affected by several factors, e.g., user preference or information topics. Among them, the competition between items could be an important reason. Usually, people in modern life may be confronted with huge amount of information of any kind. However, they could digest only a little of them due to limited attention capacities. Therefore, users may only select those they like the most, or the most significant information to read, which leads to the fierce competition among information publishers to attract attention.

 

Indeed, prior works have already concerned this phenomenon of information overload which exceeds the attention capacity of users. For instance, as shown in Fig. 1, more than 81% of users of Weibo follow more than 100 others, which leads to hundreds of new tweets every day, while most of them will be ignored. Similar situations could be found during the propagation of video, news and memes, in which popularity of each item will be degraded when a number of competing items are simultaneously available.

 

 

Along this line, two main approaches were designed for describing the propagation process in social networks. On the one hand, the mainstream of propagation modeling neglected the competition among items, for both the propagation/influence maximization problem and the retweet prediction task. One possible reason may be that it is almost impossible to differentiate which items were missed or disliked from the item set that a user received but didn′t retweet when the real propagation data was given. On the other hand, in a few agent based models, which result from the analysis on the propagation process, a user′s memory and attention capacity were considered.

 

However, traditional methods may fail to fully reflect the competition among items. Specifically, the settings of current agent based models are different from the real situation, and they could not be combined with the main stream machine learning based methods.

 

To that end, in this paper, we propose a novel algorithm to simulate the competition and present a way to model it in the main stream methods. Specifically, the scenario we considered is the propagation of one or several items, when they are spreading, there are huge amount of other items which are also diffusing. If the users on the propagating paths of items we considered receive other items, they will compete for the limited attention of these users. And we use the term “propagation background” to represent the items other than the ones being considered and will define it formally in Section 3.

 

In this study, we concentrate on the impact of propagation background and ignore other factors except the influence from neighbors. For that we cannot extract ground truth from real propagation data, we need to construct the propagation background itself first. Thus we propose a novel propagation background simulating (PBS) algorithm, which can simulate the propagation background of online social networks. In the algorithm, content items are generated by each user, the items a user could deal with at a time is limited, and a user will not conduct activities at every step. Then, we present an ICPB (independent cascade with propagation background) model to relieve the impact of propagation background, which can estimate the content item′s diffusion scope more accurately. The model has the same framework with IC model, but before a node decides whether to retweet an item, there is a probability associated with its in-degree to miss the item.

 

Specifically, our main contributions can be summarized as follows:

1) We propose the PBS algorithm which can simulate the propagation background of online social networks. The algorithm has closer settings to real scenarios than previous agent based models.

2) We present the ICPB model that relieves the propagation background′s impact on the diffusion process by improving IC model. Thus, it provides an approach to incorporate the effect of attention competition into existing propagation models.

3) We test our methods on a series of synthetic and real networks and the results demonstrate their effectiveness.

 

The remainder of this paper is organized as follows. In Section 2, we briefly discuss some related works. Section 3 presents our methods, including PBS algorithm and IC model. Next, we report and analyze the experimental results in Section 4. Finally, we conclude this paper in Section 5.

 

Download full text:

The Propagation Background in Social Networks: Simulating and Modeling

Kai Li, Tong Xu, Shuai Feng, Li-Sheng Qiao, Hua-Wei Shen, Tian-Yang Lv, Xue-Qi Cheng, En-Hong Chen

https://link.springer.com/article/10.1007/s11633-020-1227-2

http://www.ijac.net/en/article/doi/10.1007/s11633-020-1227-2

 

◇◇◇◇◇◇

For more up-to-date information:

1) WeChat: IJAC

2) Twitter:IJAC_Journal

3) Facebook:International Journal of Automation and Computing

4) Linkedin: Int.J. of Automation and Computing

5) Sina Weibo:IJAC-国际自动化与计算杂志


  • Share: