Microblog User Behaviors Analysing And Cultural Differences Mining

With the emergency of social computing research field, analysis of individual and group behaviour patterns using massive users' behaviours data in social media not only provides a new opportunity to solve problems of sociology which traditional methods cannot solve, but also are vital to practical applications, i.e., improving the design of social media function, monitoring the public events or analysing public opinion evolution. Online social network provides users with an online platform to extend the scope of their social activities, and thus users are able to share information, communicate with people from all over the world and make recommendations. As more and more users from different countries begin to use social networks, it is fairly significant to develop new features and language tools for globalized social network application platform. Moreover, the big data of global users' interaction and behaviour in online social networks also plays a key role in revealing society change and thereby to predict the development trends for society and individuals.In recent years, with the far-reaching influence of big data, there are increasingly arising researches on social computing research. However, there are only a few researches about the cultural differences behind social interaction and collaboration process of people in the online system. User behaviours, in fact, not only related to human own interest and personality, but also highly related to the cultural background of national region. In this dissertation, we analysed user behaviours in microblog and designed a behaviour model with cultural differences dimensions. From this model, we interpreted user behaviour patterns from culture's perspective. The main contributions of this dissertation are summarized as follows:1. Modeling and analysing user behaviors. Based on the large-scale integration data sets of Chinese sina weibo and Twitter, we proposed a joint probabilistic model to modeling user behaviors, which combines the explicit information such as user's attributes, behaviors and relationship together with the implicit information such as user's characteristics and communities he/she belongs to. Based on that model, we first analysed and compared user behaviour characteristics from different countries in Twitter. Based on the probabilistic user behaviors model and sample data, we analysed the network features of users from 15 countries with the most number of active users in Twitter, and found that the collective behaviours of users from some countries are distinct, based on their special characteristics. For example, some countries with smaller population display higher reciprocity than others with considerably more conversations in which users involved, furthermore, it shows that social network in microblog is more hierarchical. But more users from other countries simply use weibo as a journalism and communication platform.2. The influence of cultural differences on information dissemination. The analysis of second chapter in our thesis points out that the characteristics of the network structure of users in different countries are also different. Based on this, we proposed a dynamic mock-up of information transmission based on the structure of network, which considered the transmission probability and reciprocity in the network. We analysed the influences of characteristics of network structure and national culture's on the scope of information dissemination in the connected subnet of different countries in Twitter. We found that, the information transmission scope in the network is larger when the power distance cultural dimension score is low and the scope of information transmission is also different even if the structure characteristic of network is similar but national culture is different.3. Analysis and comparison the evolution of microblogging behavior. In microblog, user microblogging behaviour and audiences are changes through years. According to the role of user in information dissemination process and the sign convention in tweets format, we proposed twitter a user clustering method based on the characteristics of weibo, named tweets characteristics-based user clustering method (TCUCM), which clustered users as endogenous, conversationalists, generalists, echoers and link feeders. Based on TCUCM, we compared user behaviour in the longitudinal timeline, to study the evolution of user behaviour. The experimental results show that the trend and degree of user group behaviour evolution in national level is different due to the influence on cultural factors in countries.4. User activity predictability based on cultural differences. User activity is an important indicator for socialized platform evaluation. Analysis shows that user activity is associated with many factors, such as offline factors, e.g., the user's actual living conditions. Offline time, and schedule, etc. would also affect online activity. However, these behaviour data cannot be directly obtained from the weibo, therefore, only using data in Twitter is difficult to accurately predict user activities. Based on the social media features of microblog, we proposed the concept of active factors, and unified the user diverse and dynamic characteristics, as well as influence factors of social relations. Based on active factor, we designed a user activeness prediction model, UAPM. Based on UAPM, we proposed a user activity prediction algorithm to predict and analyse user activity from different countries around the world under the Twitter sample data set. The experimental results show that the predictability of user activities in different countries is of great different.