Article Preview
Top1. Introduction
Nowadays, Social media data like photos, videos and text have grown at an extraordinary rate. According to Naaman (2012), as of 2009, Flickr users have shared over 4 billion images and videos on the site, Facebook users have shared a similar amount of photos each month, and YouTube users have shared 20 new hours of video content every minute. Moreover, as of late 2010 (Miller, 2010) Twitter had 175 million registered users worldwide and they produced 65 million tweets per day. Social media sites such as Flickr and Twitter have provided different platforms for a user to share different types of information. The easy use of social media has generated huge amount of information, which may appear in different ways. Some social media sites have a platform that can support textual information, while others support visual information or both textual and visual. For example, Flickr users can perform different activities such as, posting photos and marking photos as favorites. Mining these activities is needed in order to build meaningful applications.
As a result, such huge amount of data provides many opportunities to researchers including data mining and data analytics. Not only the methods and tools to organize and manage such data have become extremely important, but also methods and tools to discover hidden knowledge from such data, which can be used for a variety of application. For example, mining of user profile on social media could help to discover any missing information including user location or gender information. However, the task of developing such methods and tools is very challenging because social media data are unstructured and different from traditional data, the privacy setting and others.
Although the social media data are different from traditional data types, data mining techniques can be used for mining user information from social media data. Data mining is the process of extracting interesting patterns or knowledge from huge amount of data (Han, Kamber, & Pei, 2011), which can be used in evolving social media such as community detection, clustering, statistical analysis, classification and association rules mining (Vakali, 2012).
Recently, a research in social media mining has been reviewed by Naaman (2012) and Chelmis and Prasanna (2011). An approach for mining social multimedia data based on a number of application is highlighted by Naaman (2012). Naaman focused on two social media applications, Flickr Landmark, and Concert Sync. Another social mining approach by Chelmis and Prasanna (2011) has presented a comprehensive study of the state of the art in social media analysis. The study is based on three aspects; (1) graph-theoretic approaches, (2) applications of semantic web technologies, and (3) data mining and analytic. Different from the above two studies, we present a comprehensive study in social user mining based on data type as well as mining techniques. In this paper, the focus is on social user mining and the use of data mining techniques. In order to achieve the goal, we present current research issues and important social media tasks. In addition, we study the problem of combining textual and visual data to perform social user mining task and show that such combination may lead to better results comparing with using individual data type.
The rest of the paper is organized as follows: Section 2 defines the problem of social user mining in general whereas Section 3 highlights the most recent work related to social user mining with respect to data type. Specifically, Section 3 focuses on three aspects:(1) social user mining base on textual information, (2) social user mining base on visual information, and (3) social user mining base on both, textual information and visual information. Section 4 summarizes the social user mining in term of the effective mining techniques followed by discussion in Section 5. The conclusion is highlighted in closing the survey at Section 6.