Xiaohui Cui

Chinese Social Media Analysis for Disease Surveillance

It is reported that there are hundreds of thousands of deaths caused by seasonal flu all around the world every year. More other diseases such as chickenpox, malaria, etc. are also serious threat to peoples physical and mental health. Therefore proper techniques for disease surveillance are highly demanded. Recently, social media analysis is regarded as an efficient way to achieve this goal, which is feasible since growing number of people post their health information to social media such as blogs, personal website, etc. Previous work on social media analysis mainly focused on English materials but hardly considered Chinese materials, which hinders the use of such technique for Chinese people. In this paper, we proposed a new method of Chinese social media analysis for disease surveillance. More specifically, we compared different kinds of methods in the process of classification, and then proposed a new way to process Chinese text data. The Chinese Sina micro-blog data collected from September to December 2013 is used to validate the effectiveness of the proposed method. The results show that a high classification precision of 87.49% in average is obtained. Comparing with the data from the authority, Chinese national influenza center, we can predict the outbreak time of flu 5 days earlier

Complex data collection in large-scale RFID systems

With the advance of RFID technology and pervasive computing, a growing number of RFID devices are deployed in the surrounding environment and form large-scale RFID systems. Many applications run on top of such a system, and perform diverse and possibly conflicting data collection tasks. Existing works about RFID data collection either focus on deducing events of interest from primitive data, or scheduling the activation of readers to mitigate various of interference. The former ones assume that the primitive data have been collected already, and the later ones assume that all the readers belong to a single application whose objective is to read all the tags once. It lacks an effective way to specify the constraints in the process of data collection for multiple applications, and coordinate the readers to meet such requirements. In this paper, we proposed a specification language and a reader coordination algorithm to solve this problem. Our language can be used to specify complex constraints in data collection tasks, based on attribute selection, set relations, and temporal relations. And then a permission based data collection approach is developed for the readers to meet these constraints in a distributed way. Extensive simulation results show that the proposed approach outperforms existing approaches in terms of the execution time.

Exploiting Feature Selection Algorithm on Group Events Data Based on News Data

As traditional media and new media such as Weibo and WeChat are increasingly used, Internet has become the main supporter of thoughts of groups or individuals, which plays a key role in guidance of our daily life and society development. This study aims to investigate/propose a note feature extraction algorithm of data processing in massive news data, extracting the key words in the news and clarifying the important ones. We try to propose a revised STF algorithm and have comparisons between efficiency of various algorithms. Experiments showed that the proposed algorithm is 4% higher on the classification accuracy than other algorithms.

Model-based sensitivity analysis on aerosol optical thickness prediction

Prediction of aerosol optical thickness (AOT) is important to study worldwide climate changes. Researchers have built multiple AOT prediction models. However, few researches were focused on the validation of input attributes for AOT regression. In this paper, we proposed a support vector regression (SVR) model-based sensitivity analysis approach to order 35 MODIS input attributes according to their sensitivity to prediction outputs. Next, the attribute sensitivity orders are used for feature selection in the context of regression by removing insensitive attribute one at a time or by removing attributes whose sensitive orders are larger than number k. The experimental results based on the collocated data between MODIS and AERONET from 2009 to 2011 showed that the top 10 insensitive attributes can be screened to speed up prediction model computation with very little loss of accuracy. The results also suggested that the top sensitive attributes are the most informative attributes, requiring the highest precision for accurate AOT prediction. Thereby, our approach will be valuable for remote sensing scientists or atmospheric scientists to optimize the design precision of top sensitive attributes in scanning equipment like MODIS and therefore improve AOT retrieval accuracy.

Planar Waypoint Generation and Path Finding in Dynamic Environment

Path Planning stands in basic positions in robotics, game AI and navigation. Previous solutions generally focused on decompose the environment into grid-like map, which is large in proportion to the size and complexity of the environment, and thus the efficiency of graph construction, update and path-finding suffers. We proposed a waypoint generation method to discretize the environment into a waypoint graph. By utilizing waypoint filtering and edge sparsification, we control the size of waypoint graph to be relatively small without sacrifice path quality in path finding. Our method limits the length of edge, and thus supports fast local update for dynamic environment. Experiments show that our way can fast generate the waypoint graph from continuous environment, and update it locally and dynamically. By integrating with physics engine, we also support path finding for multiple agents in a dynamic environment.

Analysis of Erdös Collaboration Graph and the Paper Citation Network

Network science is essential in many civil and military applications due to its great help to complex systems with network-based structures. Our research is mainly about establishing network models to measure influence and impact. We establish two models to evaluate the significance of nodes for directed network(e.g:paper citation network) and undirected network(e.g:friend circle) respectively. Our models have relative accurate assessment on the influence of nodes in a specific network and are of great expansion ability and can be applied to different sets. These models are useful in solving practical problems.

Subway Fire Evacuation Simulation Model

Fire is an increasing cause of casualties in modern life. Many people are killed in fire accidents every year. The worst thing is that people can hardly detect where and when afire will occur. Thus it is difficult to know which evacuation path is effective and safe in advance. In this article, a new system is introduced to simulate evacuation plans in fire accidents in both 2D and 3D, which helps people decide where they should go when a fire breaks out. This system is implemented in agent-based modeling, which makes it more intelligent and flexible.

Automatically constructing course dependence graph based on association semantic link model

AbstractCourse dependence graph of subject can provide an important reference model for the automatic arrangement for subject teaching plan, effective online subject learning and subject resource recommendation. Nevertheless, the challenges of the course dependence graph on the automatic construction and the maintenance of its objectivity seriously restrict its popularity. Hence, this paper proposes an approach utilizing association semantic link model for automatically constructing course dependence graph. The proposed approach employs construction of the semantic link of fragment course information resources and the association mining method to build course dependence graph. The main task of the approach can be roughly divided into the extraction of semantic key terms, the knowledge representation of course semantic and subject semantic and constructing course dependence graph. The advantages of the proposed approach are that it promotes the automation of constructing course dependence graph, defending its objectivity and getting the service of the course dependence graph smarter. The experiments show that the proposed approach has rationality and validity.

Identification of Potential Collective Actions Using Enhanced Gray System Theory on Social Media

A collective action that considerably affects government management and public security, e.g., a mass demonstration, usually experiences a long development period, originating from small and uncertain variations called weak signals on social media. Researchers generally identify collective action by small changes in communication frequency, emerging key words, and sentiment. However, most studies only consider the present environment, which may not evolve into a collective action, or conduct a short-term prediction in which significant damage is already done when the collective action is identified. This paper proposes a predictive framework to identify potential collective actions, considering the future evolution as well as the present situation, and providing a reference for early decision making. In the framework, a future sign to describe events is improved and the enhanced gray system theory is used to predict the evolution of a future sign. Mentions of events surrounding the Arab Spring—using over 300000 different open-content Web sources crawled from social media in seven different languages—are analyzed, which suggests that the predictive framework can more precisely identify the weak signals of collective actions.

Waypoint graph based fast pathfinding in dynamic environment

Pathfinding is a fundamental task in many applications including robotics, computer games, and vehicle navigation. Waypoint graph is often used for pathfinding due to its advantage in specifying the space and obstacles in a region. Currently the waypoint graph based pathfinding suffers from large computation overhead and hence long latency in dynamic environment, where the location of obstacles may change. In this paper, we propose a fast approach for waypoint graph based pathfinding in such scenario. We eliminate unnecessary waypoints and edges to make the graph sparse. And then we design a prediction-based local method to handle the dynamic change in the environment. Extensive simulation has been done and the results show that the proposed approach outperforms existing approaches.