中文摘要：

互联网的飞速发展,使人们得以从更广泛的平台获取有价值的信息.“汽车之家”等论坛网站的数据能帮助企业掌握市场格局,了解自身产品的不足.文中基于“汽车之家”论坛数据,设计了基于图挖掘的社群搜索算法MDect,能够根据用户输入的车系快速定位其所在的竞争社群,实验验证了该算法寻找出的车系社群不仅结构上联系紧密,属性上也具有较高的相似性.此外,还设计了一款整合弦图、流量图、口碑词云等多种视图的可视化系统,以展示车系之间的竞争状况与优劣势原因;最后通过用户调研与案例说明可视化方法的有效性.

英文摘要：

With the rapid development of Internet,people can get increasing number of valuable data fromvarious platforms.Effective visual analysis of data from professional websites like Autohome can be used tograsp the competitive community and disadvantages of products.In this paper,we proposed an algorithmMDect based on graph mining to detect the competitive market of specific cars.Experiments demonstratethat our method achieves a good balance between structural and attribute similarities.An interactive systemcontaining chord diagram,stream and cloud tags were developed to display the competition.Eventually weillustrated the utility of visualization through user study and cases.