面向共识的群体推荐决策模型及应用研究

The recommendation system is devoted to recommend goods and services for users, which can alleviate the information overload problem in big data era. To date, the group recommendation problem which involves a group of users has become a hot topic in the field of recommendation system. It is noted that generating consensual recommendation result(s) is very important in the group recommendation problem. However, consensus-based group recommendation research is still in its infancy stage, and many issues are still left for deep studying. To this end, this project focuses on the consensus-based group recommendation problem and designs several consensus-oriented group recommendation decision-making models to help users find the consensual recommendation result(s). First, a consensus-based group recommendation decision-making model with the minimum consensus cost is established. Then, a consensus-based group recommendation decision-making model based on preference evolution is built by taking into account the preference interactions and evolution among users. Further, a consensus-based group recommendation decision-making model under the social network is proposed by considering the social relationships among users. Following this, this project designs the optimization-based strategies with minimum changes of the social network and preference information. This project can enrich the theory system of the group recommendation and the group consensus process, which can provide decision support for the design of group recommendation system (such as tourism recommendation system).

推荐系统致力于为用户推荐商品和服务，能够有效解决大数据时代用户面临的信息过载问题。当前，群体推荐问题已经成为推荐系统领域的热点研究问题。在群体推荐问题中，如何产生共识的群体推荐结果至关重要。然而，基于共识的群体推荐研究尚处于起步阶段，还存在许多值得深入研究的问题。为此，项目对群体推荐中的共识推荐问题展开研究，建立面向共识的群体推荐决策模型。首先，建立最小共识成本群体推荐决策模型。接着，通过考虑群体用户的偏好交互和演化等复杂情形，建立基于偏好演化的群体共识推荐决策模型。进一步，通过考虑群体用户之间的社会网络关系，建立社会网络环境下的群体共识推荐决策模型，并从社会网络连接和偏好信息调控两个角度设计群体共识推荐优化策略。项目成果是对现有群体推荐和群体共识过程研究体系的丰富和完善，也能为实际群体推荐系统（例如，旅游推荐系统）的设计提供决策支持。

群体推荐致力于为群体用户推荐产品、服务或者解决方案，能够有效解决大数据时代用户面临的信息过载问题。在群体推荐问题中，如何产生共识的推荐结果至关重要。群体推荐涉及了多个研究领域，其中，群体(共识)决策理论和方法作为基本研究工具被广泛使用。项目围绕群体推荐问题中对共识推荐的现实需要，从理论层面构建了一系列群体共识决策模型。项目的主要内容包括：（1）最小偏好信息损失群体共识决策模型的建立；（2）社会信任网络下群体共识决策优化策略的设计；（3）理论结果的应用。首先，针对不同的偏好环境，建立了一系列最小偏好信息损失共识决策模型，包括：带有语言分布信息的共识决策模型，基于个性化语义的共识决策模型，异质环境下基于最小冒犯准则的分类共识决策模型等。接着，设计了关于共识决策效率的评价指标体系，构建了基于多阶段优化策略的共识决策模型。然后，考虑到社会信任关系对群体共识决策结果的影响，构建了社会信任关系驱动的最小信息损失共识决策模型，并进一步设计了带有偏好交互和演化机制的共识决策模型。最后，讨论了群体共识决策模型在可靠性管理和质量功能展开等问题中的应用。此外，设计了详细的仿真和对比实验，其结果显示了所提方法的有效性。