一种面向个性化服务的客户端细粒度用户建模方法

用户建模是实现个性化服务的关键技术。本文分析了二类用户建模存在的问题，给出了细粒度用户模型的定义，结合用户的背景知识，提出了一种客户端细粒度用户建模方法 。采用词频方法选择的特征子集和改进的k近邻分类器构成用户模型。本文的细粒度用户建模方法不需要用户的频繁交互．也不必对用户兴趣作推测，具有更好的系统亲和力
和性能。实验表明，当特征个数为40时，构建的细粒度用户模型的分类精度可达90％以上；在细粒度用户模型中，大量的特征对用户模型没有意义。

A Client-Side Fine-Grained User Modeling Method for Personalized Services

User modeling is the key technology in implementing personalized services. After an analysis of the two-class user modeling, the definition of fine-grained user model is given. Combined with the users' background, a fine-grained user modeling method is presented. A feature subset selected based on term frequency (TF) and the improved kNN classifier are used to construct a user model. The presented fine-grained user modeling method requires no frequent interaction and conclusion on training examples, which makes the system friendlier. Experiments show that the classification precision of the fine-grained user model can be over 90% with the feature number of 40. Experiments also indicate that only a small part of features contribute to user models.