基于概率主题模型的服务组合主题挖掘方法
首发时间:2016-05-18
摘要:服务个体和服务组合构成了服务系统的基本组成部分。面对数量众多、功能复杂的服务组合,如何挖掘其隐含的主题信息,对开发者创建新的服务组合的工作具有重要的指导意义。传统的主题挖掘方法一般是基于文本文档的。本文从服务组合的历史记录出发,使用概率主题模型对服务组合调用服务个体的过程进行建模,并挖掘隐含的主题。基于实际数据集ProgrammableWeb.com,本文将提出的基于服务组合历史记录的概率主题模型与传统的基于文本描述的模型进行实验对比,通过困惑度和DBI指标说明其能够挖掘出更高质量的隐含主题。
For information in English, please click here
Mining Mashup Topics based on Probabilistic Topic Model
Abstract:Service system composes of services and mashups. When dealing with mashups of different kinds and functionalities, it is significant to mine the information about latent mashup topics, which is helpful for developers when creating new mashups. Conventional topic mining methods usually focus on word documents. Different from them, in this paper, mashup usage records are utilized to reveal the latent topics with a probabilistic topic model. Based on the real-world dataset Programmable.com, the approach proposed in this paper is compared with content-based conventional method in perplexity and DBI index, illustrating that utilizing mashup usage records to build a probabilistic model could reveal latent topics of higher quality.
Keywords: probabilistic topic model mashup mashup topic
论文图表:
引用
No.4687917114718614****
同行评议
共计0人参与
勘误表
基于概率主题模型的服务组合主题挖掘方法
评论
全部评论0/1000