国内大数据赋能的智慧推荐资源管理技术研究进展

Research Advances in Resource Management Technology of Smart Recommendation Enabled by Big Data in China

[目的/意义]为资源化数据并支持智慧推荐服务多样化、提升用户交互体验,系统分析国内大数据赋能的用户兴趣建模、项目信息管理技术研究成果,阐述智慧推荐中用户及项目信息规范管理的技术体系。[方法/过程]文章用文献研究法归纳了507篇文献内容,从用户兴趣建模、项目信息管理两方面对国内大数据赋能的智慧推荐资源管理技术进行总结分析:前者包括模型表示、初始化、进化,后者包括数据采集、挖掘、存储、更新、解释。[结果/结论]大数据赋能的用户兴趣建模技术核心研究依托大数据技术在线、离线赋能用户兴趣数据分类采集及预处理、模型进化;大数据赋能的用户、项目信息管理技术核心研究用大数据技术赋能数据分类采集,并用大数据计算基础设施赋能数据离线挖掘、数据存储、数据更新、数据在线解释。现有研究需强化用户、数据双驱型的智慧推荐方案设计、实践研究,且需强化数据安全研究。

[Purpose/Significance]Data empowerment for collaborative transformation of the entire industrial chain has become a national strategy.In the context of multi-source and heterogeneous big data,we face higher demands on the diversity and real-time of recommendation services,as well as the standardization and comprehensiveness of user and item information management,in order to turn data into resources,support the diversification of smart recommendation services,and enhance the user interaction experience.To this end,we should systematically sort out and analyze the research results of user interest modeling and item information management technologies in China's big data empowerment,elaborate the technical system for standardized management of user and item information in intelligent recommendation.[Method/Process]The article summarizes the content of 507 documents using literature review methods,and summarizes and analyzes the domestic big data empowerment resource management technology for smart recommendation from two aspects:user interest modeling and item information management.User interest modeling technology includes model representation,model initialization and model evolution,the item information management technology shows the goals,implementation methods,implementation techniques,and methods of item information management,including data collection,data mining,data storage,data updating,and data interpretation.[Results/Conclusions]First,the user interest modeling technology of big data empowerment mainly studies the following issues.1)The method of model representation is to accurately formalize the user's needs.2)The model is initialized to collect user information in an implicit or explicit way and select tools and technologies according to data types to pre-process user data according to a fixed process,and select measurement methods to quantify user interest.3)Model evolution aims to apply methods and technologies to evolve user interest models offline and online.Second,the item information management technology of big data empowerment mainly studies the following issues.1)The online technology and offline equipment are used to synchronously collect item data across the whole domain,integrate domain knowledge,and use algorithm libraries and data analysis methods to mine item data.2)The distributed storage is used such as databases and file systems to store item data.3)Data collection,mining,and storage technologies are the main techniques,item data are updated through offline and online methods,and visualization technology is used to intuitively interpret ietm data.Finally,existing research needs to strengthen the design and practical research of user-driven and data-driven smart recommendation solutions,as well as strengthen research on data security.