Abstract: The component composition of Internetware became difficult in the open，dynamic and uncertain platforms such as the Internet.An optimization method on component composition is proposed for domain feature clustering.In it，Ontology classification and similarity measure are introduced，and a clustering algorithm based on partition is designed which can achieve accurate component clustering on the basis of domain feature.Attribute values of QoS about component and link are normalized，and the algorithm based on dynamic programming is proposed to solve global optimization problem of component composition.The algorithmic analysis and experiment prove that this method is effective and feasible.

Key words: component composition, domain feature, clustering dynamic programming Peer-to-Peer（P2P）

摘要： Internet环境的开放、动态和难控等特点，使网构软件的构件组装问题变得十分复杂。提出了一种面向领域特征聚类的构件组装优化方法。通过引入本体分类和相似度比较方法，设计了一种基于划分的聚类算法，以实现基于领域特征的精确的构件聚类。通过对构件和链路的多维QoS指标的换算，给出了应用动态规划方法求解面向领域特征簇的构件组装全局最优解的算法实现。算法分析和实验仿真表明了该方法的有效性和可行性。

关键词: 构件组装, 领域特征, 聚类, 动态规划, P2P