Feature Analyzing Of Social Network Users Based On Machine Learning

The growing popularity of social networks in the era of Web2.0has brought massive data to be mined. Diversity of the social networks and technologies have made it really necessary to train and learn from the data samples from social networks.Due to the rapid development of social networks, the traditional network research methods, works are no longer applicable. When analyzing users of the social network, most studies mainly focus on a particular aspect of the user, such as behavioral characteristics based on the time interval, etc. Therefore, this study conducted a comprehensive analysis of the individual user’s attributes, information, and feature. Furthermore, unsupervised learning algorithm—spectral clustering is adopted and achieved upon data collected from Weibo.Given that social network analysis is multidisciplinary, the study uses methods such as machine learning, text-processing tools to research into Weibo user’s information propagation characteristics, behavior characteristics, attributes characteristics, relationship characteristics, text feature. Visualization is an inevitable trend in the era of big data; this study has achieved the visualization of network characteristics.Work completed in this study includes:social networking features analysis based on graph theory, complex network, classical virus propagation model; visualization of these network features is achieved by Gephi; modified model SEIR is used to describe the propagation of tweets in Weibo; A model-Interests Driven Interaction Model is established to portray user behavior in Weibo. Furthermore, LDA (Latent Dirichlet Allocation) is chosen to produce theme distribution for actual users in Weibo; on this basis, similarity between users is evaluated; and spectral clustering is conducted to mine the community so that users in the same community would share similar interests. Through theoretical and empirical analysis, this paper comprehensively analyzes information, attributes, relationships, behavioral characteristics of user in Weibo; satisfactory results is achieved when spectral clustering is applied to learn the data samples from Weibo.