一种基于隐语义概率模型的个性化Web服务推荐方法

发布时间：2018-01-19 04:00

本文关键词： Web服务推荐个性化隐语义概率模型指标偏好期望极大算法　出处：《计算机研究与发展》2014年08期 　论文类型：期刊论文

【摘要】：为了满足Web服务使用者的个性化需求,提出了一种基于隐语义概率模型的用户指标偏好预测方法,用于个性化Web服务推荐.首先,引入两个决定用户指标偏好的关键因素:用户以及用户所处的服务情境,隐语义概率模型借助隐含类别建立用户指标偏好、用户及服务情境三者之间的隐含语义依赖关系,并且为描述用户、服务情境、指标偏好多方面的特征,允许这三者可同时以不同的概率隶属于多个隐含类别;然后,将期望极大(expectation maximization,EM)算法运用于由层次分析法获得的训练数据,以估计隐语义概率模型的参数;最后,使用该模型预测用户在特定服务情境下的指标偏好.隐语义概率模型与标准的基于内存的协同过滤以及基于聚类改进的协同过滤相比,不仅具有明确的数学模型,而且实验结果表明,隐语义概率模型对用户个性化指标偏好的预测精度最高,同时可以缓解数据稀疏性带来的不良影响.
[Abstract]:In order to meet the personalized needs of Web service consumers, a user preference prediction method based on implicit semantic probability model is proposed for personalized Web service recommendation. This paper introduces two key factors to determine user index preference: the user and the service situation in which the user is located. The implicit semantic probability model establishes the user index preference with the help of the implicit category. The implicit semantic dependency between the user and the service situation, and to describe the characteristics of user, service situation and index preference, which allows them to belong to multiple hidden categories with different probability at the same time. Then, the expected maximization algorithm is applied to the training data obtained by AHP to estimate the parameters of the hidden semantic probability model. Finally, the model is used to predict the index preference of users in a specific service situation. The hidden semantic probability model is compared with the standard memory-based collaborative filtering and the improved cooperative filtering based on clustering. The experimental results show that the implicit semantic probabilistic model has the highest prediction accuracy to the user's personalized index preference and can alleviate the adverse effect of data sparsity.
【作者单位】：中国科学院软件研究所天基综合信息系统重点实验室;中国科学院大学;
【基金】：国家“八六三”高技术研究发展计划基金项目(2012AA011206)
【分类号】：TP393.09;TP391.3
【正文快照】： 随着互联网技术的飞速发展,面向服务的计算(service-oriented computing,SOC)[1]作为一种新的计算模式得到了广泛应用.Web服务就是这样一种分布运行于Internet之上、支持不同平台之间互操作的松耦合软件系统.它主要通过“发布-查找-绑定”的模式允许服务使用者和提供者之间形

【相似文献】