伴随着移动互联网的蓬勃发展，越来越多的数据被存储在互联网中，这些海量的数据给我们带来丰富选择的同时也带来了信息过载的问题。为了解决这一问题，推荐系统应运而生。推荐系统能够依据用户和项目的交互历史自动构建它们之间的联系，帮助用户在大量信息中快速地找到他们感兴趣的内容或产品。学术领域和工业界的研究表明，推荐系统在解决信息过载问题、提高数据利用率方面表现出良好的效果。然而，传统的推荐算法却受到附属信息和模型结构的限制，难以达到预期效果。此外，数据稀疏问题尤其是冷启动问题也常常影响推荐算法的准确性。当推荐系统中新用户或新项目缺乏交互历史时，传统的推荐算法往往无法学习到新用户或新项目的信息，从而难以为用户提供准确的推荐结果，这就是所谓的冷启动问题，是数据稀疏问题的一种极端表现。

为了解决上述问题，本文将知识图谱引入到推荐系统中，研究使用深度学习方法来提取知识图谱中丰富的语义信息进行辅助推荐，主要工作如下：

基于协同信号的知识图注意力网络推荐算法（The Collaborative signal Knowledge Graph Attention Network for Recommender algorithm，CKGAN），首先CKGAN使用用户已交互过的历史项目作为用户的初始表示，根据这些项目和候选项目在知识图谱上找到与其对应的实体以及与这些项目相关的实体。然后使用改进图注意力网络来融合相关实体的信息得到具体的项目向量。最后为了解决用户可能存在多种兴趣，在预测层中，面对不同的候选集动态地建模了用户的向量表示。通过在三个真实的公共数据集上与主流的推荐算法进行对比实验，结果表明，CKGAN在CTR预测和top-K推荐中都有显著提升。

基于知识图谱的多神经网络序列推荐算法（The Multi-neural network of Sequential recommendation algorithm for Knowledge Graph，MSKG），为了挖掘出用户与项目交互背后的兴趣以及用户兴趣的变化趋势，本文提出通过多种神经网络模型来训练序列推荐模型。MSKG首先通过使用知识图谱嵌入的方法来学习用户顺序的历史数据中每个项目的向量表示，然后将这些项目划分到多个时间窗口中，通过采用自注意力机制挖掘出用户的局部兴趣，再让其通过双向长短期记忆网络，刻画出用户的兴趣变化。最后将学习到的用户偏好向量和待推荐项目向量通过多层感知机来预测用户下一次的交互是否为待推荐项目。本文通过在两个真实数据集与多个序列推荐算法进行对比实验，在HR@K和NDCG@K两个评价指标上的结果表明，模型能够更加精准的预测目标用户下一步行动。

With the rapid development of mobile Internet, more and more data is stored on the Internet, which brings rich choices to us, but also brings the problem of information overload. To solve this problem, recommender systems have emerged. Recommender systems can automatically build connections between users and items based on their interaction history, helping users quickly find the content or products they are interested in from a large amount of information. Research in the academic and industrial fields has shown that recommender systems perform well in solving the problem of information overload and improving data utilization. However, traditional recommendation algorithms are limited by auxiliary information and model structures, making it difficult to achieve the expected results. In addition, the data sparsity problem, especially the cold start problem, often affects the accuracy of recommendation algorithms. When there is no interaction history for new users or new items in the recommendation system, traditional recommendation algorithms often cannot learn information about new users or new items, making it difficult to provide accurate recommendation results for users, which is called the cold start problem and is an extreme manifestation of the data sparsity problem.

To solve the above problems, this paper proposes knowledge graph into recommender systems and studies the use of deep learning methods to extract rich semantic information from knowledge graph for auxiliary recommendation. The main work is as follows:

The Collaborative signal Knowledge Graph Attention Network for Recommender algorithm (CKGAN) is based on the collaborative signal and knowledge graph attention network, which uses the historical projects that users have interacted with as the initial representation of users. Based on these projects and candidate projects, CKGAN finds the corresponding entities and entities related to these projects on the knowledge graph. Then, an improved graph attention network is used to fuse the information of related entities to obtain specific project vectors. Finally, to address the possibility of multiple user interests, the user vector representation is dynamically modeled in the prediction layer for different candidate sets. Through comparative experiments with mainstream recommendation algorithms on three real public datasets, the results show that CKGAN significantly improves both CTR prediction and top-K recommendation.

The Multi-Neural Network Sequence Recommendation Algorithm Based on Knowledge Graphs (MSKG), in order to mine the interest behind the interaction between users and items and the changing trend of user interest, this paper proposes to train the sequence recommendation model through a variety of neural network models. MSKG first uses the knowledge map embedding method to learn the vector representation of each item in the user's sequential historical data, and then divides these items into multiple time windows, and uses the self-attention mechanism to mine the user's local interests, so that It uses a two-way long and short-term memory network to describe the user's interest changes. Finally, the learned user preference vector and the vector of the recommended item are predicted using a multi-layer perceptron to predict whether the user's next interaction will be with the recommended item. Through comparative experiments with multiple sequence recommendation algorithms on two real datasets, the results show that MSKG can more accurately predict the target user's next action in terms of HR@K and NDCG@K evaluation metrics.

[1]Bai C, Yan H, Yin S, et al. Exploring the development trend of internet finance in China: Perspective from club convergence[J]. The North American Journal of Economics and Finance, 2021, 58(1): 1-12.

[2]Guo Q, Zhuang F, Qin C, et al. A Survey on Knowledge Graph-Based Recommender Systems [J]. IEEE Transactions on Knowledge & Data Engineering, 2020, 34(08): 3549-3568.

[3]Hall M, Frank E, Holmes G, et al. The WEKA data mining software: an update[J]. ACM SIGKDD explorations newsletter, 2009, 11(1): 10-18.

[4]陈旭松. 基于用户行为序列建模的推荐算法研究[D].中国科学技术大学,2021.

[5]Sarwar B, Karypis G, Konstan J, et al. Item-based collaborative filtering recommendation algorithms [C] //Proceedings of the 10th international conference on World Wide Web. Switzerland: WWW, 2001: 285-295.

[6]Rivas A, Chamoso P, González-Briones A, et al. Social network recommender system, a neural network approach[C] //Intelligent Data Engineering and Automated Learning–IDEAL 2020. Guimaraes: IDEAL, 2020: 213-222.

[7]Pan Y, Huo Y, Tang J, et al. Exploiting relational tag expansion for dynamic user profile in a tag-aware ranking recommender system[J]. Information Sciences, 2021, 545(1): 448-464.

[8]胡琪, 朱定局, 吴惠粦等. 智能推荐系统研究综述[J]. 计算机系统应用, 2022, 31(04): 47-58.

[9]Pazzani M J, Billsus D. Content-based recommendation systems[J]. The adaptive web: methods and strategies of web personalization, 2007(1): 325-341.

[10]Koren Y, Bell R, Volinsky C. Matrix factorization techniques for recommender systems[J]. Computer, 2009, 42(8): 30-37.

[11]Yao L, Sheng Q Z, Ngu A H H, et al. Exploring recommendations in internet of things[C] //Proceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval. New York: ACM, 2014: 855-858.

[12]Rendle S. Factorization machines[C] //2010 IEEE International conference on data mining. Sydney: IEEE, 2010: 995-1000.

[13]Guo H, Tang R, Ye Y, et al. DeepFM: a factorization-machine based neural network for CTR prediction[C] //Proceedings of the 26th International Joint Conference on Artificial Intelligence. Melbourne: AAAI, 2017: 1725-1731.

[14]Wang R, Shivanna R, Cheng D, et al. Dcn v2: Improved deep & cross network and practical lessons for web-scale learning to rank systems[C] //Proceedings of the web conference 2021. Ljubljana: ACM, 2021: 1785-1797.

[15]He X, Liao L, Zhang H, et al. Neural collaborative filtering [C] //Proceedings of the 26th international conference on World Wide Web. Switzerland: ACM, 2017: 173-182.

[16]Hong M, Koo C, Chung N. DSER: Deep-Sequential Embedding for single domain Recommendation[J]. Expert Systems with Applications, 2022, 208(1): 1-13.

[17]He X, Du X, Wang X, et al. Outer product-based neural collaborative filtering[C] //Proceedings of the 27th International Joint Conference on Artificial Intelligence. Stockholm: AAAI, 2018: 2227-2233.

[18]葛尧, 陈松灿. 面向推荐系统的图卷积网络[J]. 软件学报, 2020, 31(04): 1101-1112.

[19]Quadrana M, Cremonesi P, Jannach D. Sequence-aware recommender systems[J]. ACM Computing Surveys (CSUR), 2018, 51(4): 1-36.

[20]Sanna Passino F, Maystre L, Moor D, et al. Where to next? A dynamic model of user preferences[C] //Proceedings of the Web Conference 2021. Ljubljana: ACM, 2021: 3210-3220.

[21]Zhang Y, Dai H, Xu C, et al. Sequential click prediction for sponsored search with recurrent neural networks[C] //Proceedings of the AAAI Conference on Artificial Intelligence. Québec City: AAAI, 2014, 28(1): 1369-1375.

[22]Hidasi B, Karatzoglou A. Recurrent neural networks with top-k gains for session-based recommendations[C] //Proceedings of the 27th ACM international conference on information and knowledge management. Torino: ACM, 2018: 843-852.

[23]Tang J, Wang K. Personalized top-n sequential recommendation via convolutional sequence embedding[C] //Proceedings of the eleventh ACM international conference on web search and data mining. Marina: ACM, 2018: 565-573.

[24]Kang W C, McAuley J. Self-attentive sequential recommendation[C] //2018 IEEE international conference on data mining (ICDM). Singapore: IEEE, 2018: 197-206.

[25]Zhou G, Mou N, Fan Y, et al. Deep interest evolution network for click-through rate prediction[C] //Proceedings of the AAAI conference on artificial intelligence. Honolulu: AAAI, 2019, 33(01): 5941-5948.

[26]Feng Y, Lv F, Shen W, et al. Deep session interest network for click-through rate prediction[C] //Proceedings of the 28th International Joint Conference on Artificial Intelligence. Macao: IJCAI, 2019: 2301-2307.

[27]Ji S, Pan S, Cambria E, et al. A survey on knowledge graphs: Representation, acquisition, and applications[J]. IEEE transactions on neural networks and learning systems, 2021, 33(2): 494-514.

[28]Guo Q, Zhuang F, Qin C, et al. A survey on knowledge graph-based recommender systems[J]. IEEE Transactions on Knowledge and Data Engineering, 2020, 34(8): 3549-3568.

[29]Zhang F, Yuan N J, Lian D, et al. Collaborative knowledge base embedding for recommender systems[C] //Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. San Francisco: ACM, 2016: 353-362.

[30]Lin Y, Liu Z, Sun M, et al. Learning entity and relation embeddings for knowledge graph completion [C] //Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence. Palo Alto: AAAI, 2015: 2181-2187.

[31]Zhang Y, Ai Q, Chen X, et al. Learning over knowledge-base embeddings for recommendation[J]. Algorithms, 2018, 11(9): 137.

[32]Bordes A , Usunier N , Garcia-Duran A , et al. Translating Embeddings for Modeling Multi-relational Data[C] //Neural Information Processing Systems. Curran Associates California: NeurlPS 2013: 1-8.

[33]Wang H, Zhang F, Xie X, et al. DKN: Deep knowledge-aware network for news recommendation[C] //Proceedings of the 2018 world wide web conference. Lyon: ACM, 2018: 1835-1844.

[34]Ji G, He S, Xu L, et al. Knowledge graph embedding via dynamic mapping matrix[C] //Proceedings of the 53rd annual meeting of the association for computational linguistics and the 7th international joint conference on natural language processing. Beijing:ACL, 2015: 687-696.

[35]Mikolov T, Chen K, Corrado G, et al. Efficient estimation of word representations in vector space[J]. Computer Science, 2013(1)：1-10.

[36]Cao Y, Wang X, He X, et al. Unifying knowledge graph learning and recommendation: Towards a better understanding of user preferences[C] //The world wide web conference. San Francisco: ACM, 2019: 151-161.

[37]Wang Z, Zhang J, Feng J, et al. Knowledge graph embedding by translating on hyperplanes[C] //Proceedings of the AAAI conference on artificial intelligence. Québec City: AAAI, 2014, 28(1).

[38]Wang H, Zhang F, Zhao M, et al. Multi-task feature learning for knowledge graph enhanced recommendation[C] //The world wide web conference. San Francisco: ACM, 2019: 2000-2010.

[39]Yu X, Ren X, Gu Q, et al. Collaborative filtering with entity similarity regularization in heterogeneous information networks[J]. IJCAI HINA, 2013, 27.

[40]Luo C, Pang W, Wang Z, et al. Hete-cf: Social-based collaborative filtering recommendation using heterogeneous relations[C] //2014 IEEE International Conference on Data Mining. Shenzhen: IEEE, 2014: 917-922.

[41]Shi C, Zhang Z, Ji Y, et al. SemRec: a personalized semantic recommendation method based on weighted heterogeneous information networks[J]. World Wide Web, 2019, 22: 153-184.

[42]Sun Z, Yang J, Zhang J, et al. Recurrent knowledge graph embedding for effective recommendation[C] //Proceedings of the 12th ACM conference on recommender systems. Vancouver: ACM, 2018: 297-305.

[43]Wang X, Wang D, Xu C, et al. Explainable reasoning over knowledge graphs for recommendation[C] //Proceedings of the AAAI conference on artificial intelligence. Hawaii: AAAI, 2019, 33(01): 5329-5336.

[44]Wang H, Zhang F, Wang J, et al. Ripplenet: Propagating user preferences on the knowledge graph for recommender systems[C] //Proceedings of the 27th ACM international conference on information and knowledge management. Torino:ACM, 2018: 417-426.

[45]Nickel M, Tresp V, Kriegel H P. A three-way model for collective learning on multi-relational data[C] // The 28th International Conference on Machine Learning. Washington ICML. 2011: 1-8.

[46]Wang H, Zhao M, Xie X, et al. Knowledge graph convolutional networks for recommender systems [C] //The World Wide Web conference. New York: ACM, 2019: 3307-3313.

[47]Wang H, Zhang F, Zhang M, et al. Knowledge-aware graph neural networks with label smoothness regularization for recommender systems[C] //Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining. Alaska: ACM, 2019: 968-977.

[48]Kipf T N , Welling M. Semi-Supervised Classification with Graph Convolutional Networks [C] //the 5th International Conference on Learning Representations. San Diego ICLR, 2017: 1-14.

[49]Tang X, Wang T, Yang H, et al. AKUPM: Attention-enhanced knowledge-aware user preference model for recommendation[C]// Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. Alaska: ACM, 2019: 1891-1899.

[50]Wang X, He X, Cao Y, et al. Kgat: Knowledge graph attention network for recommendation [C]// Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining. New York: ACM, 2019: 950-958.

[51]Zhang D, Liu L, Wei Q, et al. Neighborhood aggregation collaborative filtering based on knowledge graph[J]. Applied Sciences, 2020, 10(11): 3818-3832.

[52]Hui B, Zhang L, Zhou X, et al. Personalized recommendation system based on knowledge embedding and historical behavior[J]. Applied Intelligence, 2022, 52(1): 954-966.

[53]Xu Z, Liu H, Li J, et al. CKGAT: Collaborative Knowledge-Aware Graph Attention Network for Top-N Recommendation[J]. Applied Sciences, 2022, 12(3): 1669-1631.

[54]Guo J, Zhou Y, Zhang P, et al. Trust-aware recommendation based on heterogeneous multi-relational graphs fusion[J]. Information Fusion, 2021, 74(1): 87-95.

[55]Dong X, Gabrilovich E, Heitz G, et al. Knowledge vault: A web-scale approach to probabilistic knowledge fusion[C] //Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. New York: ACM, 2014: 601-610.

[56]Shuman D I, Narang S K, Frossard P, et al. The emerging field of signal processing on graphs: Extending high-dimensional data analysis to networks and other irregular domains[J]. IEEE signal processing magazine, 2013, 30(3): 83-98.

[57]Defferrard M, Bresson X, Vandergheynst P. Convolutional neural networks on graphs with fast localized spectral filtering[C] //Proceedings of the 29th Advances in Neural Information Processing Systems. Barcelona: NeurIPS, 2016: 3844-3852.

[58]Velickovic P, Cucurull G, Casanova A, et al. Graph attention networks [C] //Proceedings of the 6th International Conference on Learning Representations. Vancouver: ICLR, 2018: 1-12.

[59]Rendle S, Freudenthaler C, Gantner Z, et al. BPR: Bayesian personalized ranking from implicit feedback[C] //Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence. Arlington: AUAI , 2009: 452-461.

[60]Wang H, Yao K, Luo J, et al. An implicit preference-aware sequential recommendation method based on knowledge graph[J]. Wireless Communications and Mobile Computing, 2021(1): 1-12.