基于用户模型的个性化广告推荐技术研究

发布时间：2018-10-11 06:36

【摘要】：随着互联网的快速普及和智能携带产品的推广,越来越多的人投入到互联网中,互联网成为现代广告业务的新载体。互联网广告投放具有效益高、覆盖面广等特点,因此受到前所未有的关注。传统的互联网广告投放充满着随机性和不确定性,导致用户在上网的时候面临着铺天盖地的广告,降低了用户的上网体验,造成了广告位的转化率低等问题。为了提升用户体验,研究者提出了个性化广告推荐技术,并成为了近年来的研究热点。个性化广告推荐技术的核心是用户模型,个性化广告服务的质量与用户模型的精确性有直接相关性。本文提出了基于显隐式信息结合的用户模型的个性化广告推荐技术。开展了如下研究工作:(1)提出一种改进的隐式建模方法。传统的隐式建模方法将用户所有的日志信息作为建模信息。改进的隐式建模方法对用户上网日志进行分析,将查询词与历史文档进行相似性对比,过滤掉相似性值较小的文档,提高用户兴趣模型精确度。(2)提出了显隐式信息相结合的用户建模技术。显隐式信息结合的用户建模首先是根据用户提交的用户信息初始化用户模型,然后通过对用户上网历史信息进行分析,构建隐式用户模型,对初始用户模型进行更新。(3)提出了一种基于用户模型的协同过滤广告推荐算法。协同过滤技术是根据用户-项目评分矩阵,挖掘出用户相似集合,通过近邻集合中兴趣相近的用户给目标用户推荐信息。本文将上面提出的用户兴趣模型应用到协同过滤算法中,利用用户模型矩阵替代评分矩阵,实验结果表明基于用户模型的协同过滤推荐算法能够提升广告推荐的精确度。
[Abstract]:With the rapid popularization of the Internet and the promotion of intelligent carrier products, more and more people put into the Internet, the Internet has become a new carrier of modern advertising business. Internet advertising has the characteristics of high efficiency and wide coverage, so it has received unprecedented attention. The traditional Internet advertising is full of randomness and uncertainty, which results in the users facing numerous advertisements when they surf the Internet, which reduces the users' experience on the Internet and causes the low conversion rate of advertisements. In order to improve the user experience, the researchers put forward the personalized advertising recommendation technology, which has become a research hotspot in recent years. The core of personalized advertising recommendation technology is user model. The quality of personalized advertising service is directly related to the accuracy of user model. In this paper, a personalized advertising recommendation technology based on explicit and implicit information combination is proposed. The main works are as follows: (1) an improved implicit modeling method is proposed. The traditional implicit modeling method takes all user log information as modeling information. The improved implicit modeling method analyzes the user log, compares the query words with the historical documents, and filters out the documents with small similarity value. Improve the accuracy of user interest model. (2) A user modeling technique combining explicit and implicit information is proposed. Firstly, the user model is initialized according to the user information submitted by the user, and then the implicit user model is constructed by analyzing the historical information of the user. The initial user model is updated. (3) A collaborative filtering advertising recommendation algorithm based on user model is proposed. Collaborative filtering technology is based on the user-item scoring matrix to mine the similar set of users and recommend information to the target users through the users with similar interests in the nearest neighbor set. In this paper, the user interest model is applied to the collaborative filtering algorithm, and the user model matrix is used to replace the scoring matrix. The experimental results show that the collaborative filtering recommendation algorithm based on the user model can improve the accuracy of advertising recommendation.
【学位授予单位】：湖南工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】