面向大数据的群组推荐理论与方法

While recommender systems suggest items that the individual users might like, group recommender systems (GRS) are designed to operate in contexts in which more than one person is involved in the recommendation process. The group-oriented approach is enabled to lower the computational complexity, while the return of investment on recommender systems is increased. However, GRSs are in early stages of development, and face the following issues: i) inefficient group detection, ii) undetermined preferences extraction and iii) ineffective preference aggregation. This project is expected to simplify group detection through unified dimension reduction and clustering, quantify the preference offset for improving the deterministic of extraction, and comprehensively represent the group preference through weight-based multidimensional aggregation. The main purpose of this project is to build a theoretical foundation and develop key techniques for big-data-oriented GRS. Specifically, it is expected to propose a key-factor-based approach for group detection, design a sentiment-analysis-based approach for extracting determined preference and develop a weight-based multidimensional preference aggregation. This project will further promote the development of GRS in the age of big data, and provide a novel approach to improve the efficiency and accuracy of recommender systems.

群组推荐系统将服务对象由个体用户扩展为多个用户，不仅可以缓解推荐系统的计算复杂度，还能有效地提高使用推荐系统的投入产出比。但是群组推荐系统的研究尚处于起步阶段，理论基础尚不完备，存在潜在群组发现方法效率有限、获取用户偏好存在不确定性、群组偏好融合策略缺乏有效性等问题。本项目拟通过探索通用的降维和聚类方法，简化群组发现的计算复杂度；研究针对用户偏好不确定性的误差量化方法，提高用户偏好获取结果的精准性；构建基于加权模型的多维融合策略，全面有效地描述群组偏好。本项目的总体目标是建立面向大数据的群组推荐系统的若干理论基础和关键技术，提出基于关键因子的潜在群组发现理论与方法，提出基于情感分析的用户偏好不确定性处理机制，提出基于加权模型的多维偏好融合策略。预期成果不仅为进一步推动群组推荐技术在大数据环境下的广泛应用打下坚实基础，也为提高推荐系统的数据处理能力和用户偏好获取的准确性提供了新的研究思路。

针对当前群组推荐系统存在的关键问题：（1）潜在群组发现方法效率有限；（2）获取用户偏好存在不确定性；（3）群组偏好融合策略缺乏有效性。本项目借助机器学习、数据挖掘等技术，提高潜在群组发现的效率，解决用户偏好的不确定性，构建多维偏好融合策略。具体地，探索通用的降维和聚类方法，简化群组发现的计算复杂度；研究针对用户偏好不确定性的误差量化方法，提高用户偏好获取结果的精准性；构建基于加权模型的多维融合策略，全面有效地描述群组偏好。通过实验验证和模拟仿真，证实了本项目所提出的群组推荐理论与方法，可以有效提高群组推荐系统的高效性和精确性，提高了群组推荐系统的鲁棒性，降低了群组偏好融合的复杂度。为推动群组推荐系统走向实际应用提供理论支撑和技术支持，具有重要的研究意义和广阔的应用背景。