面向目标市场的信息最大覆盖算法

发布时间：2018-05-08 12:34

本文选题：社会网络 + 目标营销　；参考：《计算机学报》2014年04期

【摘要】：当一个企业或商家需要投放广告时,往往会先通过历史数据、个人兴趣等挖掘出可能购买自己产品的用户,即目标市场(Target Market),然后将广告信息通过电视、报纸等公共媒体的形式传递给这些目标用户,希望有更多的目标用户接受信息.然而调查显示,相比于传统大众媒体,人们更倾向于从自己认识的人那里去获取信息,因此文中考虑利用社会影响力的方式去传播广告:在社会网络中说服有限数目的初始用户,并让他们向熟识的人传播信息,期望信息可以通过级联传播覆盖尽可能多的目标用户.由于以往的信息覆盖最大化的工作集中于对全局网络的考虑,因此会忽略目标节点和全局网络之间的联系.通过数据观察可以发现,目标用户往往会由于同质性等原因而聚集在一起,因此文中提出基于聚类的KCC算法,算法通过对用户进行聚类分析,找出每个聚类的代表性用户,使得这些代表性节点可以影响尽可能多的目标用户,同时避免他们之间对信息覆盖的重叠.在不同的真实的数据集的实验显示KCC可以在大多数情况下取得优于其它常用算法的性能,尤其当种子节点数增多时,KCC可以更多地避免节点之间信息覆盖的重叠,从而取得更好的效果;同时,KCC只需要很短的运行时间,具有良好的可扩展性.
[Abstract]:When an enterprise or business needs to place an ad, it often uses historical data, personal interest, and so on to find out who might buy its own product, the target market, and then send the advertising message over television. The form of public media, such as newspapers, is passed on to these target users in the hope that more target users will receive information. However, surveys show that people tend to get information from people they know more than traditional mass media. So the article considers using social influence to spread advertising: persuading a limited number of initial users in social networks and allowing them to spread information to people they know well. The desired information can be propagated in cascades to cover as many target users as possible. Since the previous work of maximizing information coverage focused on the consideration of the global network, the relationship between the target node and the global network would be ignored. Through the observation of data, it can be found that the target users tend to gather together because of homogeneity and other reasons. Therefore, a clustering based KCC algorithm is proposed in this paper. By clustering analysis, the algorithm finds out the representative users of each cluster. These representative nodes can affect as many target users as possible and avoid overlapping of information coverage between them. Experiments on different real data sets show that KCC can achieve better performance than other commonly used algorithms in most cases, especially when the number of seed nodes increases, it can avoid the overlap of information coverage between nodes. In addition, KCC only needs very short running time and has good expansibility.
【作者单位】：南京大学计算机软件新技术国家重点实验室;南京陆军指挥学院军队管理系;
【基金】：国家自然科学基金委创新研究群体科学基金(61321491) 国家自然科学基金(61202113) 国家自然科学基金重大集成项目基金(91218302) 江苏省自然科学基金(BK2011510)资助~~
【分类号】：TP301.6;G206

【共引文献】