摘要: 本贴列出实验室在推荐系统方向的一些工作. 主要目的是保存 bib 方便以后查阅与引用.
1. SCI 期刊论文
1.1 基于随机森林的三支推荐
- 三支推荐开山之作, 基于隐式评价 (即是否购买).
- 将用户、项目、评分三个表格融合, 构建相应的决策表.
- 引用次数 200+.
@article{ZhangMin2016Three,
title = {Three-way recommender systems based on random forests},
author = {Heng-Ru Zhang and Fan Min},
journal = {Knowledge-Based Systems},
year = {2016},
volume = {91},
pages = {275--286},
doi = {10.1016/j.knosys.2015.06.019}
}
1.2 基于回归的三支推荐
- ZhangMin2016Three 的姊妹篇, 基于显式评价 (即评分).
- 两个阈值独立地学习. 测试集的最优阈值与训练集的一致.
- 也是高被引.
@article{ZhangMin2017Regression,
title = {Regression-based three-way recommendation},
author = {Heng-Ru Zhang and Fan Min and Bing Shi},
journal = {Information Sciences},
year = {2017},
volume = {378},
pages = {444--461},
doi = {10.1016/j.ins.2016.03.019}
}
1.3 三支交互混合推荐
- 在三支推荐的基础上, 引入多轮会话, 更好地模拟实际推荐场景.
- 在不同的阶段使用不同的推荐方案: 冷启动时使用流行项目, 有用户评分后更精准地推荐. 结合 exploit 与 explore 结合. 前者保证推荐列表中有用户喜欢的项目, 后者探索用户其它可能喜欢的项目.
@article{Xu2022Hybrid,
author = {Yuan-Yuan Xu and Shen-Ming Gu and Hua-Xiong Li and Fan Min},
title = {A hybrid approach to three-way conversational recommendation},
journal = {Soft Computing},
year = {2022},
volume = {26},
pages = {13885--13897},
doi = {10.1007/s00500-022-07416-x}
}
1.4 启发式概念格推荐方法
- 创建概念集合支持协同过滤.
- 不需要创建概念格, 因此高效.
- 是这个方向我们第一篇 SCI, 有一定开创性. 折腾了三年.
@article{Liu2022Heuristic,
author = {Zhong-Hui Liu and Qi Zhao and Lu Zou and Wei-Hua Xu and Fan Min},
title = {A heuristic concept construction approach to collaborative recommendation},
journal = {International Journal of Approximate Reasoning},
year = {2022},
volume = {146},
pages = {119--132},
doi = {10.1016/j.ijar.2022.04.004}
}
1.5 在正态分布假设下的推荐系统魔法边界估计
- 魔法边界是指学习器能够达到的最佳性能, 相当于分类器的 Bayes 误差.
- 魔法边界对评估数据质量, 指导推荐算法的设计方面具有很强的理论意义.
- 很少有人做这个方向, 因为多数工作致力于提升推荐性能 (更准确).
- 设计了三个正态分布模型来估计.
- 这篇论文从有想法到发表经历了 3 年以上.
引用次数 5
@article{Zhang2018Magic,
author = {Heng-Ru Zhang and Fan Min and Yan-Xue Wu and Zhuo-Lin Fu and Lei Gao},
title = {Magic barrier estimation models for recommended systems under normal distribution},
journal = {Applied Intelligence},
year = {2018},
volume = {48},
number = {12},
pages = {4678--4693},
doi = {10.1007/s10489-018-1237-8}
}
1.5 在混合高斯假设下的推荐系统魔法边界估计
- 在 Zhang2018Magic 基础上的工作.
- MoG 比正态分布更加贴合实际.
- 引用次数 1
@article{Zhang2022Mixture,
author = {Heng-Ru Zhang and Jie Qian and Hui-Lin Qu and Fan Min},
title = {A Mixture-of-Gaussians model for estimating the magic barrier of the recommender system},
journal = {Applied Soft Computing},
year = {2022},
volume = {114},
pages = {108162},
doi = {10.1016/j.asoc.2021.108162}
}
1.6 去除离群点以提高训练数据的质量
@article{Xu2022Improving,
author = {Yuan-Yuan Xu and Shen-Ming Gu and Fan Min},
title = {Improving recommendation quality through outlier removal},
journal = {International Journal of Machine Learning and Cybernetics},
year = {2022},
volume = {13},
number = {7},
pages = {1819--1832},
doi = {10.1007/s13042-021-01490-7}
}
1.7 局部信息嵌入的增强slope one推荐算法
@article{Zhang2019Esli,
author = {Heng-Ru Zhang and Yuan-Yuan Ma and Xin-Chao Yu and Fan Min},
title = {ESLI: Enhancing slope one recommendation through local information embedding},
journal = {Plos one},
year = {2019},
volume = {14},
number = {10},
pages = {e0222702},
doi = {10.1371/journal.pone.0222702}
}
1.8 基于情感评论的矩阵分解 SBMF+R
- 将带有情感的评论引入矩阵分解模型, 提高推荐的可靠性.
- 萍萍是从化工方向转到机器学习, 博士期间又转到管理, 最后在那边的次顶刊发表了论文.
- 引用次数 47
@article{Shen2019Sentiment,
author = {Rong-Ping Shen and Heng-Ru Zhang and Hong Yu and Fan Min},
title = {Sentiment based matrix factorization with reliability for recommendation},
journal = {Expert Systems with Applications},
year = {2019},
volume = {135},
pages = {249--258},
doi = {10.1016/j.eswa.2019.06.001}
}
1.9 多通道特征向量支持推荐
- 提出新的相似性度量–多通道特征向量 (MCFV), 并利用该度量设计了新的推荐算法
引用次数 25
@article{Zhang2019Efficient,
author = {Heng-Ru Zhang and Fan Min and Zhi-Heng Zhang and Song Wang},
title = {Efficient collaborative filtering recommendations with multi-channel feature vectors},
journal = {International Journal of Machine Learning and Cybernetics},
year = {2019},
volume = {10},
number = {5},
pages = {1165--1172},
doi = {10.1007/s13042-018-0795-8}
}
1.10 提出推荐算法集成的适应性机制(AMRE)。AMRE由三部分组成,包括一组代理、一个奖励函数和一个轮询机制。
引用次数 1
@article{Huang2019Adaptive,
author = {Qi Huang and Yuan-Yuan Xu and Yong Chen and Heng-Ru Zhang and Fan Min},
title = {An adaptive mechanism for recommendation algorithm Ensemble},
journal = {IEEE Access},
year = {2019},
volume = {7},
pages = {10331--10342},
doi = {10.1109/ACCESS.2019.2891561}
}
1.12 结合三角和Jaccard相似度,提出了一种新的推荐方法
引用次数 56
@article{Sun2017Integrating,
author = {Shuang-Bo Sun and Zhi-Heng Zhang and Xin-Ling Dong and Heng-Ru Zhang and Tong-Jun Li and Lin Zhang and Fan Min},
title = {Integrating Triangle and Jaccard similarities for recommendation},
journal = {PloS one},
year = {2017},
volume = {12},
number = {8},
pages = {1--16},
doi = {10.1371/journal.pone.0183570}
}
2. 国际会议论文
3. 核心期刊论文
3.1 提出了一种融合直接和间接的用户非对称信任关系的推荐算法(ATRec)
引用次数 3
@article{Zhang2018Recommendation,
author = {Zi-Yin Zhang and Heng-Ru Zhang and Yuan-Yuan Xu and Qin Qin},
title = {Recommendation algorithm combining user's asymmetric trust relationships},
journal = {Computer Science},
year = {2018},
volume = {45},
number = {10},
pages = {37--42},
doi = {10.11896/j.issn.1002-137X.2018.10.007}
}
3.2 为解决评分系统中存在的数据稀疏问题,提出一种内涵粗糙三支概念以及基于它的个性化推荐方案
引用次数 0
@article{Liu2022Three,
author = {Zhong-Hui Liu and Xin Li and Fan Min},
title = {Three-wayconcept with rough intent and its application in personalized recommendation},
journal = {Journal of Northwest University(Natural Science Edition)},
year = {2022},
pages = {1--10},
doi = {10.16152/j.cnki.xdxbzr.2022-05-005}
}
3.3 介绍了三元概念的构建,并将应用于社会化推荐
引用次数 1
@article{Liu2021Heuristic
author = {Zhong-Hui Liu and Qi Zhao and Lu Zou and Fan Min},
title = {Heuristic construction of triadic concept and its application in social recommendation},
journal = {Computer Science},
year = {2021},
pages = {234--240},
issue = {06},
doi = {10.11896/jsjkx.200500136}
}
3.4 提出一种基于启发式概念构造的组推荐方法
引用次数 3
@article{Liu2019Group,
author = {Zhong-Hui Liu and Lu Zou and Mei Yang and Fan Min},
title = {Group recommendation with concept of heuristic construction},
journal = {Journal of Frontiers of Computer Science and Technology},
year = {2019}
volume = {14},
number = {4},
pages = {703--711},
doi = {10.3778/j.issn.1673-9418.1905012}
}
3.5 提出一种可解释性泛化矩阵分解推荐算法
引用次数 0
@article{Lv2022An,
author = {Ya-Lan Lv and Yuan-Yuan Xu and Heng-Ru Zhang},
title = {An explainable generalized matrix factorization recommendation algorithm},
journal = {Journal of Nanjing University(Natural Sciences)},
year = {2022},
volume = {58},
number = {1},
pages = {135--142},
doi = {10.13232/j.cnki.jnju.2022.01.013}
}
3.6 为同时提高推荐系统的准确度和效率, 提出了一种多通道特征向量的新三角距离推荐算法
引用次数 0
@article{Lv2021Efficient,
author = {Ya-Lan Lv and Heng-Ru Zhang and Qin Qin and Yuan-Yuan Xu},
title = {Efficient Recommendation of New Triangular Distance for Multi-channel Feature Vectors},
journal = {Journal of Suthwest University(Natural Science)},
year = {2021},
volume = {43},
number = {10},
pages = {19--28},
doi = {10.13718/j.cnki.xdzk.2021.10.003}
}
3.7 针对油气生产中的抽油机井参数优化问题,提出了一种基于抽油机井生产调控、维护措施数据的抽油机井生产参数优化的粒计算方法
引用次数 1
@article{Zhang2020Granular,
author = {Heng-Ru Zhang and Ke-Lin Zhu and Yuan-Yuan Xu and Ying Qiao},
title = {Granular computing for pumping well parameter optimization},
journal = {Journal of Southwest Petroleum University (Science & Technology Edition)},
year = {2020},
volume = {42},
number = {6},
pages = {115},
doi = {10.11885/j.issn.1674-5086.2020.05.29.02}
}
3.8 提出基于信任传递机制的三支推荐算法
引用次数 4
@article{Qin2020Three,
author = {Qin Qin and Heng-Ru Zhang},
title = {Three-way recommendation based on trust transfer mechanism},
journal = {Pattern Recognition and Artificial Intelligence},
year = {2020},
volume = {33},
number = {7},
pages = {600--609},
doi = {10.16451/j.cnki.issn1003-6059.202007003}
}
3.9 模拟现实电子商务推荐场景,设计三支交互推荐模型,提出结合流行度区间和 M-distance k 近邻的混合推荐算法
引用次数 6
@article{Xu2019Three,
author = {Yuan-Yuan Xu and Heng-Ru Zhang and Fan Min and Yu-Ting Huang},
title = {Three-way interactive recommendation},
journal = {Journal of Nanjing University (Natural Sciences)},
year = {2019},
volume = {55},
number = {6},
pages = {973--983},
doi = {10.13232/j.cnki.jnju.2019.06.010}
}