享乐性与实用性商品的个性化推荐策略与技术研究——基于用户评论大数据驱动的商品特征挖掘

Existing recommendation techniques are not able to differentiate between hedonic and utilitarian products effectively, so that they cannot grasp users’ experiential preference and utilitarian requirement precisely. Such drawback may lead to a large number of redundant recommendations, which lowers the effectiveness of recommender systems. This project brings the relevant concepts of hedonic and utilitarian products from marketing science into recommendation domain, and focuses on how to make use of the hedonic and utilitarian dimensions of products for improving the quality of recommendation, which has not been studied by existing research. Design Science Research methodology will be adopted as a guideline. By improving and adapting data crawling, text analysis and data mining techniques, this project will implement precise extraction of the experiential and utilitarian features of products from user reviews. Furthermore, the project will carry out research on the recommendation strategies and techniques for hedonic products, utilitarian products and hybrid products respectively, with the objective of improving the accuracy, diversity and interpretability of the recommendations. The research findings of this project will not only enrich the theories of recommender systems and precise marketing, but also provide practical suggestions for the development and improvement of recommender systems of a variety of E-Commerce websites, which will bring economic benefits to the companies.

现有的推荐技术没有对享乐性商品和实用性商品进行有效区分，导致无法准确把握用户的体验偏好和功能需求，容易产生大量的冗余推荐，限制了推荐系统有效性的充分发挥。本课题把营销学中有关享乐性商品与实用性商品的相关概念引入个性化推荐领域，聚焦如何有效利用商品的享乐性和实用性提高推荐质量这一现有研究尚未触及的问题。本课题将以设计科学研究方法为指导，通过对数据抓取技术、文本分析技术和数据挖掘技术的改进和创新，实现从用户评论大数据中精准地提取出商品的体验特征和功能特征，并对享乐性商品、实用性商品及两者并存时的推荐策略和相关技术展开研究，以实现推荐精度、推荐多样性和推荐可解释性的显著提高。本课题的研究成果将对推荐系统和精准营销等相关领域的理论进行有效补充，同时可为各类电子商务网站的推荐系统开发或改进提供可操作的建议，为企业带来经济效益。

现有的推荐技术没有对享乐性商品和实用性商品进行有效区分，导致无法准确把握用户的体验偏好和功能需求，容易产生大量的冗余推荐，限制了推荐系统有效性的充分发挥。本课题把营销学中有关享乐性商品与实用性商品的相关概念引入个性化推荐领域，聚焦如何有效利用商品的享乐性和实用性提高推荐质量这一现有研究尚未触及的问题。本课题以设计科学研究方法为指导，构建了有效的特征抽取技术和综合推荐方案，并利用电影、图书、移动应用和婚恋等领域的实际数据进行验证，结果表明所提出的技术方案可以有效提高推荐的准确性、多样性、覆盖率和可解释性。本课题在国际主流期刊发表了一系列研究成果，对推荐系统和精准营销等相关领域的理论进行有效补充，同时可为各类电子商务网站的推荐系统开发或改进提供可操作的建议，为企业带来经济效益。