Guozhong Dong

WiFi-Based Real-Time Calibration-Free Passive Human Motion Detection †

With the rapid development of WLAN technology, wireless device-free passive human detection becomes a newly-developing technique and holds more potential to worldwide and ubiquitous smart applications. Recently, indoor fine-grained device-free passive human motion detection based on the PHY layer information is rapidly developed. Previous wireless device-free passive human detection systems either rely on deploying specialized systems with dense transmitter-receiver links or elaborate off-line training process, which blocks rapid deployment and weakens system robustness. In the paper, we explore to research a novel fine-grained real-time calibration-free device-free passive human motion via physical layer information, which is independent of indoor scenarios and needs no prior-calibration and normal profile. We investigate sensitivities of amplitude and phase to human motion, and discover that phase feature is more sensitive to human motion, especially to slow human motion. Aiming at lightweight and robust device-free passive human motion detection, we develop two novel and practical schemes: short-term averaged variance ratio (SVR) and long-term averaged variance ratio (LVR). We realize system design with commercial WiFi devices and evaluate it in typical multipath-rich indoor scenarios. As demonstrated in the experiments, our approach can achieve a high detection rate and low false positive rate.

Discovering burst patterns of burst topic in twitter

A burst topic user graph model is proposed.Hierarchical clustering is applied to cluster burst topics and reveal burst patterns from the macro perspective.Frequent sub-graph mining is used to discover information flow patterns of burst topic from the micro perspective. Twitter has become one of largest social networks for users to broadcast burst topics. There have been many studies on how to detect burst topics. However, mining burst patterns in burst topics has not been solved by the existing works. In this paper, we investigate the problem of mining burst patterns of burst topic in Twitter. A burst topic user graph model is proposed, which can represent the topology structure of burst topic propagation across a large number of Twitter users. Based on the model, hierarchical clustering is applied to cluster burst topics and reveal burst patterns from the macro perspective. Frequent sub-graph mining is used to discover the information flow patterns of burst topic from the micro perspective. Experimental results show that several interesting burst patterns are discovered, which can reveal different burst topic clusters and frequent information flows of burst topic. Display Omitted

Detecting Anomalies in Microblogging via Nonnegative Matrix Tri-Factorization

With the increasing of anomalous user’s intelligent, it is difficult to detect the anomalous users and messages in microblogging. Most of the studies attempt to detect anomalous users or messages individually nowadays. In this paper, we propose a co-clustering algorithm based on nonnegative matrix tri-factorization to detect anomalous users and messages simultaneously. A bipartite graph between user and message is built to model the homogeneous and heterogeneous interactions, and homogeneous relations as constraints to improve the accuracy of heterogeneous co-clustering algorithm. The experimental results show that the proposed algorithm can detect anomalous users and messages with high accuracy on Sina Weibo dataset.

Information Diffusion and Influence Measurement Based on Interaction in Microblogging

In microblogging, user interaction is the main factor that promotes the information diffusion rapidly. According to the user interaction in the process of information diffusion, this paper proposes a directed tree model based on user interaction that considering the history, type and frequency of interaction. User interaction matrix was used to describe the interactions between pairs of users. A directed diffusion tree was generated from the sparsification of interaction graph. The edges of directed diffusion tree were used to measure the information influence and identify the spam in microblogging. Experimental results show that the directed tree model can describe the information diffusion, measure the influence more accurately and identify the spam in the dataset more effectively.

Detecting Bots in Follower Markets

Along with the widely using of microblog, third party services such as follower markets sell bots to customers to build fake influence and reputation. However, the bots and the customers that have large numbers of followers usually post spam messages such as promoted messages, messages containing malicious links. In this paper, we propose an effective approach for bots detection based on interaction graph model and BP neural network. We build an interaction graph model based on user interaction and design robust interaction-based features. We conduct a comprehensive set of experiments to evaluate the proposed features using different machine learning classifiers. The results of our evaluation experiments show that BP neural network classifier using our proposed features can be effectively used to detect bots compared to other existing state-of-the-art approaches.

DEPO: Detecting Events of Public Opinion in Microblog

The rapid spread of microblog messages and sensitivity of unexpected events make microblog become the center of public opinion. Because of the large amount of microblog message stream and irregular language of microblog message, it is important to detect events of public opinion in microblog. In this paper, we propose DEPO, a system for Detecting Events of Public Opinion in microblog. In DEPO, abnormal messages detection algorithm is used to detect abnormal messages in the real-time microblog message stream. Combined with EPO (Events of Public Opinion) features, each abnormal message can be formalized as EPO features using microblog-oriented keywords extraction method. An online incremental clustering algorithm is proposed to cluster abnormal messages and detect EPO.

Detecting Community Pacemakers of Burst Topic in Twitter

Twitter has become one of largest social networks for users to broadcast burst topics. Influential users usually have a large number of followers and play an important role in the diffusion of burst topic. There have been many studies on how to detect influential users. However, traditional influential users detection approaches have largely ignored influential users in user community. In this paper, we investigate the problem of detecting community pacemakers. Community pacemakers are defined as the influential users that promote early diffusion in the user community of burst topic. To solve this problem, we present DCPBT, a framework that can detect community pacemakers in burst topics. In DCPBT, a burst topic user graph model is proposed, which can represent the topology structure of burst topic propagation across a large number of Twitter users. Based on the model, a user community detection algorithm based on random walk is applied to discover user community. For large-scale user community, we propose a ranking method to detect community pacemakers in each large-scale user community. To test our framework, we conduct the framework over Twitter burst topic detection system. Experimental results show that our method is more effective to detect the users that influence other users and promote early diffusion in the early stages of burst topic.

Heterogeneous Sparse Relational Data Co-clustering in Social Network

In social network, users generated multi-typed entities and complex interactive relations. The relational data mining is a hot research in social computing. Co-clustering algorithms have been proposed to mine underlying structure of different entities in heterogeneous social network. However, the real heterogeneous relational data are very sparse. In this paper, we propose a fast High-order Sparse Non-negative Matrix Factorization algorithm to co-cluster heterogeneous sparse relational data based on Correlation Matrix(HSNMF-CM), which is built by the correlation relations of small entities. In HSNMF-CM, the sparseness and size of matrix are reduced simultaneously. Under the sparse constraint, the block coordinate descent algorithms are used to accelerate the convergence rate of the matrix factorization. We assess the performance of the HSNMF-CM on two social data sets. The results show that our algorithm outperforms state-of-the-art algorithms on accuracy and convergence speed, and possesses a high scalability on large-scale heterogeneous relational data sets.

SAMD: A System for Abnormal Messages Detection Oriented Microblog Message Stream

Microblog has been an important medium for providing the rapid communications of public opinion and can quickly publicize a burst topic for discussion when unexpected incidents happen. Abnormal messages are usually the source of burst topics and are important for the diffusion of burst topics. It is necessary to detect abnormal messages from microblog real-time message stream. In this paper, we propose SAMD, a System for Abnormal Messages Detection. In SAMD, sliding time window model is applied to divide the microblog data stream into different shards. Only that the participation of messages exceed initial threshold can be indexed and stored in two-level hash table. An efficient abnormal messages detection model is used to detect abnormal messages in a given time window. The case study on the collected data set can show that SAMD is effective to detect and demonstrate abnormal messages from large-scale microblog message stream.

Agglomerative Hierarchical Co-clustering Based on Bregman Divergence

Recently, co-clustering algorithms are widely used in heterogeneous information networks mining, and the distance metric is still a challenging problem. Bregman divergence is used to measure the distance in traditional co-clustering algorithms, but the hierarchical structure and the feature of the entity itself are not considered. In this paper, an agglomerative hierarchical co-clustering algorithm based on Bregman divergence is proposed to learn hierarchical structure of multiple entities simultaneously. In the aggregation process, the cost of merging two co-clusters is measured by a monotonic Bregman function, integrating heterogeneous relations and features of entities. The robustness of algorithms based on different divergences is tested on synthetic data sets. Experiments on the DBLP data sets show that our algorithm improves the accuracy over existing co-clustering algorithms.