推荐算法研究进展  

Survey on Research Progress of Recommendation Algorithms

作　　者：李孟浩 赵学健[1] 余云峰 宋学永 孙知信[1] LI Meng-hao;ZHAO Xue-jian;YU Yun-feng;SONG Xue-yong;SUN Zhi-xin(School of Modern Posts,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;Guoji Beisheng(Nanjing)Technology Development Co.,Ltd,Nanjing 210001,China)

机构地区：[1]南京邮电大学现代邮政学院,南京210003 [2]国基北盛(南京)科技发展有限公司,南京210001

出　　处：《小型微型计算机系统》2022年第3期544-554,共11页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61972208)资助;中国博士后科学基金项目(2018M640509)资助。

摘　　要：随着互联网的发展,全球数据量爆炸式增长,信息过载现象严重,如何获取用户真正关注的信息成为困扰人们的难题之一.在此背景下,推荐算法在各个领域得到了广泛应用.本文首先介绍了目前主流推荐算法的分类方法和主要评价指标.其次,分别介绍了当前各类推荐算法的研究进展,其中包括传统推荐算法的基础推荐原理和研究进展以及神经网络在推荐算法中的研究应用,对其进行归纳总结.同时分析了数据稀疏性、冷启动和可伸缩性等推荐算法常见问题.最后,提出了现有推荐算法的不足以及在应用中遇到的部分问题,介绍了未来推荐算法的研究热点.With the development of the Internet,the amount of global data has exploded,and information overload is serious.How to obtain the information you really want has become a long-term problem that plagues people.In this background,recommendation algorithms have been widely used in various fields.This article first introduces the classification and important evaluation indexes of related recommendation algorithms.This article introduces the current research progress of recommendation algorithms in each category,including the basic recommendation principles and research progress of traditional recommendation algorithms and the application of neural networks in recommendation algorithms,and summarizes them in the end.At the same time,common problems of recommendation algorithms such as data sparsity,cold start and scalability are analyzed.Finally,the shortcomings of existing recommendation algorithms and some problems encountered in the application are proposed,and the research hotspots of future recommendation algorithms are introduced.

关 键 词：推荐算法 评价指标 神经网络 数据稀疏性 冷启动 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]