无线泛在环境下用户业务行为预测分析.pdfVIP

摘要 各种终端以及多种类型网络的并存，形成了重叠覆盖的无线泛在环境。在无线泛在环境 下，为了提升用户体验质量，为用户推送个性化定制业务，有必要学习、了解用户的业务需 求行为及规律，以实现用户业务行为的准确预测。因此本文针对无线泛在环境下用户业务行 为预测进行了深入的研究，主要工作如下： 首先，对无线泛在环境的相关理论进行介绍；在此基础上，分析了无线泛在环境下用户 行为及其研究现状；进一步地，详述了本文相关的理论工具。 其次，提出了一种基于多维度Markov 模型的用户业务行为预测算法。在构建Markov 模 型的过程中，将用户业务时长这一因素考虑在内，增加时间维度，然后通过用户业务访问倾 向权重调整概率转移矩阵，最后利用构造出的Markov 模型来对用户未来的业务偏好进行预 测，并通过仿真验证了该算法的有效性。 最后，提出了一种基于Markov 和关联规则的用户业务行为预测算法。考虑到Markov 模 型的缺陷，低阶的Markov 模型仅依据最近的历史状态作出预测，准确率难以保障；而高阶 的Markov 模型需要较长的历史状态序列，导致预测样本量过小，难以用于预测，故考虑将 Markov 模型和关联规则算法相结合，提出了一种新颖的用户业务行为预测算法，并通过仿真 验证了该算法的有效性。 关键词 无线泛在环境，用户行为，预测算法，: Markov模型，关联规则算法 I 万方数据 Abstract The coexistence and overlapping of terminals and networks lead to wireless ubiquitous environment. In wireless ubiquitous environment, in order to improve user quality of experience and provide personalized customization services for users, it is essential to study user service demands as well as user behavior laws, and predict user service behavior accurately. Therefore ,it is necessary to study user service behavior in wireless ubiquitous environment. The main works are as follow: Firstly, The related theory of wireless ubiquitous environment is introduced in this thesis and then the related researches of user behavior and its current research situation are surveyed. Further the related theoretical theory tools are introduced in detail. Secondly, Based on multi-dimension markov mode, a new kind of prediction algorithm is proposed. In the process of constructing markov model, the length of user service time is taken into account. Then the probability transfer matrix of markov model is adjusted according to the user service access tendency w