COLLECTIVE SOCIAL ACTIVITY PREDICTION BY USING CONTINUOUS-TIME STOCHASTIC PROCESS

  • : Ms Word, Ms Word Format
  • : 70 Pages
  • : ₦3,000 | $25 | ₵60 | Ksh 2720
  • : 1-5 Chapters
  •  
  • Click to DOWNLOAD Materials

COLLECTIVE SOCIAL ACTIVITY PREDICTION BY USING CONTINUOUS-TIME STOCHASTIC PROCESS

Abstract

In recent years, online social network sites have successfully emerged to attract a huge number of users. Due to the tremendous number of users, these sites are playing a more and more important role in business marketing and customer care. The ex-ante knowledge of the future social activity and network status can assist a lot in decision making and improve the profit of businesses.

Existing approaches analyze social networks and predict their future behaviors by modeling the evolving network with all cumulative nodes (members) and connections. Although the arrival of new nodes is taken into consideration to address the dynamic nature of social networks, in general, the degree of fine-grained node activity (e.g. a user adds a friend or posts on a friend’s wall) has played very little role in such an analysis. As the infrastructure of a social network heavily depends on the member activities, we propose to explore the collective social dynamics and community infrastructure to predict the future social activity. As the evolution of social networks is a random process, we make use of a continuous-time stochastic process model with an uncountable future state space to simulate the social dynamics and investigate the impact of user activities.

In this thesis, we derive a novel parameterized model that incorporates the information embedded in collective activities to predict various features of the social community, including the size of active population, the number of social interactions, the social sentiment trend, etc. With member activities evolving over time, the predicting model itself also evolves and therefore dynamically simulates the network status to fit the real-time characteristics of the current active population. Our experiments using two real social media datasets (Facebook and CiteSeer) show that the proposed parameterized stochastic model is effective to simulate the social activity evolution and can predict the collective social activities with more than 80% accuracy through different time scales.

 

Chapter 1

Introduction

1.1        Social Network Analysis

The intensive and highly diversified communications among people lead to complex social network infrastructure. With the emergence of online social networking sites and online communities, a lot of resources arise to enable in-depth social networking research towards better understanding of the complex social phenomena. Among the social networking research areas, an important topic is to understand and predict the dynamics of user behaviors and interactions, as well as the establishment and evolution of social relationship [1, 2, 3, 4, 5, 6, 7] .

In conventional social networking research, social networks are usually represented by graphs, where vertices denote social actors and edges denote social relationships. Therefore, the notion of social network evolution is mostly defined as the expansion of the social network graph. Different approaches have been proposed to analyze and predict the social network evolution and information diffusion. They can be roughly grouped into three categories: static network mining [8] , microscopic evolution prediction [9, 10] , and structural analysis [11, 12] .

In static network mining, structural patterns, such as power-law distribution and small-world phenomenon, are discovered by mining snapshots of the networks. In microscopic evolution prediction, various models are proposed to simulate the social network growth at the micro scope. In structure analysis, a variety of structural features are measured and different models are presented to explain their evolution.

However, approaches that are purely based on graph structures could be biased when social actors demonstrate different levels of activity. Highly active social actors make more contribution towards the interaction, expansion and functionality of the online society, while non-active and less-active ones only contribute as static nodes in the graph. We argue that the notion of social interactions is an important indicator in social network analysis.

1.2   A Challenge: Similar Topology but Different Activities

Example 1. Figure 1.1 shows two real-world social groups extracted from the Facebook friendship network [13] . They have similar structural evolution process in terms of friendship links, and the final states are shown in Figure 1.1 (a) and (c). Hence, they are expected to demonstrate similar behaviors (e.g., information diffusion) in pure structure-based analysis. However, social interactions (wall postings) within the two groups turn out to be quite different, as shown in Figure 1.1 (b) and (d). As we can see, members in one of the groups frequently communicate 3 on Facebook, while the other group is quite inactive.

As we can see, it is impractical to predict social network activities only using social connections (i.e. relationships). A significant portion of the social network accounts are inactive, which causes purely structure-based evolution models to be less accurate. Moreover, compared with friendship links, social activity is a more meaningful indicator of the social dynamics and user behaviors within social networks.

1.3        Various Social Activities

A lot of research is conducted on investigating the evolution of social activities. Many studies aim to capture structural properties [14, 15] , while their influence is only evaluated on the change of network structure at the individual level. For instance, in [16] , qualitative evolution patterns of co-authorship network are summarized. However, these studies are only applicable in predicting a certain type of social activity. A model that can predict the addition of new nodes and new edges is not applicable to predict the sentiment change of the social group. Therefore, we come up with a question: Is it possible to create a generic model that can simulate and predict various types of social activities?

1.4          Collective Social Activity Analysis

To address the inefficiency of social connections, we propose a new approach to predict social network activities based on the collective activity analysis.

By using a continuous-time stochastic process to simulate social activity evolution, we address both the randomness and the tendency of evolution. At the same time, the predicting model is parameterized with the current activity features; thus predictions are not based on historical average (like most existing work), but instead, determined by the inherent characteristics and real-time observation of the social network. In addition, predictions made by our continuous-time stochastic model have an uncountable future state space, i.e. numeric values; therefore our model is advantageous in quantifying social activity evolution.

Moreover, the proposed model could be applied to any arbitrary subset of users in a social network (will be elaborated later), which makes it extremely flexible. Based on the target of application, our model can be applied to any social network and predict any collective social activity feature. With intensive experiments from real-world social network data, the proposed solution is proven effective in predicting social activities.

1.5         Scope and Structure of Thesis

In this thesis, we address the challenge of social activity prediction based on collective member activities. We specifically explored the following two perspectives to solve this problem:

(1)Given that social behaviors are random and subject to environment changes, how to simulate the evolution of social activities effectively and accurately?

(2)As the social network evolves over time, the pattern of social activities also evolves. How to incorporate this change when modeling the social activity evolution?

The rest of this thesis is organized as follows. In Chapter 2, we discuss related works and formally define the problem with an overview of our solutions. In Chapter 3, we present the preliminaries of continuous-time stochastic process. Then we explain the details of developing and solving the parameterized evolution model in Chapter 4. After that, we show the experimental validation in Chapter 5. Finally, we summarize this thesis in Chapter 6.

COLLECTIVE SOCIAL ACTIVITY PREDICTION BY USING CONTINUOUS-TIME STOCHASTIC PROCESS

Sharing is caring!

Leave a Reply