中文摘要：

人群定向是广告推荐系统中的一种重要技术，它是通过分析种子人群的行为数据，找出潜在的目标人群，而现有人群定向算法大多依赖于传统的协同过滤推荐算法．由于传统的协同过滤算法具有推荐精度低和抗攻击能力较弱的问题，为了解决这些问题，提出了一种基于邻居选取策略的人群定向算法．1）通过用户行为相似，动态选择出与种子人群具有相似行为的用户；2）以用户特征和用户行为作为邻居选取的依据，通过用户相似度从行为相似人群中选择出每个种子用户的邻居，并将所有的相似邻居作为候选人群；3）通过基于邻居选取策略的人群定向算法，从候选人群中择出潜在的目标用户，以完成人群定向．实验结果表明：与现有方法相比，该方法不仅提高了人群定向的精度，而且也增强了系统的抗攻击能力．

英文摘要：

Audience targeting which is designed to discover the prospective target users by analyzing the these seed users＇ behavior is an important technology in the online advertising recommendation systems, and the existing audience targeting technologies mostly rely on collaborative filtering algorithms. However, the traditional collaborative filtering algorithms have the disadvantages of lower precision and weaker anti-attack capability. In order to solve the problems, an audience targeting algorithm based on neighbor choosing strategy is proposed. Firstly, the users which have the similar behavior with the seed audiences are chosen dynamically by means of the user behavior similarity. Then, on the basis of the users＇ feature and behavior, the neighbors of each seed user are chosen from the behavior similar audiences by the user similarity, and all the neighbors are considered to be the candidate audiences. Finally, the prospective audiences are chosen from the candidate users by the audience targeting algorithm based on neighbor choosing strategy, so as to complete the task of audience targeting. Compared with the existing methods, the experimental results on real-world advertisement datasets show that the audience targeting algorithm not only improves the precision, but enhances the anti-attack capability as well.