Haizheng Zhang

Advances in Web Mining and Web Usage Analysis

Adaptive Website Design Using Caching Algorithms.- Incorporating Usage Information into Average-Clicks Algorithm.- Nearest-Biclusters Collaborative Filtering with Constant Values.- Fast Categorization of Web Documents Represented by Graphs.- Leveraging Structural Knowledge for Hierarchically-Informed Keyword Weight Propagation in the Web.- How to Define Searching Sessions on Web Search Engines.- Incorporating Concept Hierarchies into Usage Mining Based Recommendations.- A Random-Walk Based Scoring Algorithm Applied to Recommender Engines.- Towards a Scalable kNN CF Algorithm: Exploring Effective Applications of Clustering.- Detecting Profile Injection Attacks in Collaborative Filtering: A Classification-Based Approach.- Predicting the Political Sentiment of Web Log Posts Using Supervised Machine Learning Techniques Coupled with Feature Selection.- Analysis of Web Search Engine Query Session and Clicked Documents.- Understanding Content Reuse on the Web: Static and Dynamic Analyses.

An LDA-based Community Structure Discovery Approach for Large-Scale Social Networks

Community discovery has drawn significant research interests among researchers from many disciplines for its increasing application in multiple, disparate areas, including computer science, biology, social science and so on. This paper describes an LDA(latent Dirichlet Allocation)-based hierarchical Bayesian algorithm, namely SSN-LDA (simple social network LDA). In SSN-LDA, communities are modeled as latent variables in the graphical model and defined as distributions over the social actor space. The advantage of SSN-LDA is that it only requires topological information as input. This model is evaluated on two research collaborative networkst: CtteSeer and NanoSCI. The experimental results demonstrate that this approach is promising for discovering community structures in large-scale networks.

HSN-PAM: Finding Hierarchical Probabilistic Groups from Large-Scale Networks

Real-world social networks are often hierarchical, re- flecting the fact that some communities are composed of a few smaller, sub-communities. This paper describes a hierarchical Bayesian model based scheme, namely HSN- PAM (Hierarchical Social Network-Pachinko Allocation Model), for discovering probabilistic, hierarchical com- munities in social networks. This scheme is powered by a previously developed hierarchical Bayesian model. In this scheme, communities are classified into two categories: super-communities and regular-communities. Two differ- ent network encoding approaches are explored to evaluate this scheme on research collaborative networks, including CiteSeer and NanoSCI. The experimental results demon- strate that HSN-PAM is effective for discovering hierarchi- cal community structures in large-scale social networks.

Graphical models based hierarchical probabilistic community discovery in large-scale social networks

Real-world social networks, while disparate in nature, often comprise of a set of loose clusters (a.k.a. communities), in which members are better connected to each other than to the rest of the network. In addition, such communities are often hierarchical, reflecting the fact that some communities are composed of a few smaller, sub-communities. Discovering the complicated hierarchical community structure can gain us deeper understanding about the networks and the pertaining communities. This paper describes a hierarchical Bayesian model based scheme namely hierarchical social network-pachinko allocation model (HSN-PAM), for discovering probabilistic, hierarchical communities in social networks. This scheme is powered by a previously developed hierarchical Bayesian model. In this scheme, communities are classified into two categories: super-communities and regular-communities. Two different network encoding approaches are explored to evaluate this scheme on research collaborative networks, including CiteSeer. The experimental results demonstrate that HSN-PAM is effective for discovering hierarchical community structures in large-scale social networks.

Introduction to the Special Issue on Linking Social Granularity and Functions

In recent years, social network research has advanced significantly. Researchers are increasingly interested in investigating network structure features and social patterns residing in social networks, including community discovery, link prediction, anomaly detection, trend prediction, etc. [Newman 2003]. Even though detection of these features and patterns is of great interest, understanding their functions in broader applications remains relatively limited. Encountering a real application about social networks, it is important to select and study the corresponding network structure features and social patterns of the same scale. For example, given a macro-level problem like information maximization, the macro-level structure features such as community are worth great consideration [Wu 2008], given a micro-level problem like user profiling, the micro-level structure features such as neighborhoods are important [Mislove et al. 2010]. Furthermore, network structure features and social patterns in different scales are highly correlated to each other. For example, Yin et al. [2010a, 2010b] suggest that both macro-level network structure and micro-level node attribute information could be leveraged together on the tasks of link prediction and node attribute inference. Not only are the macro-level communities often used to detect micro-level links [Zheleva and Getoor 2009], but the micro-level patterns like user moods can also be used to identify influential users of the whole network [Quercia et al. 2011]. In this special issue, we have accepted six articles that carefully link the scale of social features to the corresponding functions. Each accepted article has gone through two to three rounds of reviewing, each round with three referees. The contents of this special issue cover decreasing the granularity of network structure features for solving micro-level problems like user interest profiling and link sign prediction or improving the efficiency of time-consuming information maximization, moving beyond network clusters themselves by studying their overlap regions, which paves a contradictive way against conventional community detection, and leveraging both macro-level network structure features and micro-level node attributes for link prediction and community detection. Interestingly, all of the articles consider the selection of features in right scale as crucial to applications. A macro-level feature like network structure has been widely utilized in various data mining tasks. However, the granularity of global network structure is too coarse to calibrate the accuracy of data mining. In this special issue, there are two articles focusing on the study of designing local network structure features to boost the effectiveness of micro-level data mining problems, that is, user interest profiling and link sign prediction. The article “Infer User Interests via Link Structure Regularization” by Jinpeng Wang, Wayne Xin Zhao, Yulan He, and Xiaoming Li proposes that the network structure feature should be task-dependent. The authors suggest that graph regularization on network structure should be based on the particular task of which node similarities are evaluated in a local link structure, and they verify this assumption by learning user interest for a set of predefined topics based on a retweet network in Twitter. The article “Cluster-Based Collaborative Filtering for Sign Prediction in Social Networks with Positive and Negative Links” by Amin Javari and Mahdi Jalili proposes that