基于动态注意力强化学习的可解释学习路径推荐

doi:10.3969/j.issn.0255-8297.2026.01.008

摘要/Abstract

摘要： 大规模在线教育的普及使得学习者面临课程选择困难，个性化学习路径推荐面临依赖单一模态数据导致语义表征局限，以及静态知识图谱难以生成动态可解释推荐逻辑的挑战。为解决上述问题，提出一种基于动态注意力强化学习的可解释学习路径推荐（explainable learning path recommendation based on dynamic attention reinforcement learning,ELPR-DARL）框架。首先，构建了异构协同知识图谱，集成课程文本、视觉内容及知识依赖关系，增强跨模态语义对齐能力；其次，设计了邻接节点动态注意力聚合机制，通过偏置修正策略调整实体关系权重，并利用双向交互聚合器融合多阶邻域特征，提升知识推理的细粒度表达能力；最后，提出知识图谱感知的强化学习策略，基于路径连通性奖励函数显式建模用户行为与知识拓扑的关联，生成包含全局奖励与局部注意力权重的可解释路径。基于MOOC数据集上的实验表明，本方法在NDCG、Recall、HR和Precision指标上分别达到22.85%、33.81%、52.01%和6.34%，较次优模型提升2.88%、3.55%、2.42%和3.26%。用户调研显示，80.36%的学习者认为路径解释显著提升了推荐透明度。本研究验证了动态注意力机制与强化学习的协同优化能有效平衡推荐精度与可解释性。

关键词: 协同知识图谱, 学习路径推荐, 可解释推荐, 动态注意力机制, 强化学习, 推荐系统

Abstract: The popularization of large-scale online education has made it difficult for learners to choose courses, and personalized learning path recommendation faces the challenge of relying on single modal data, which leads to the limitation of semantic representation. Moreover, static knowledge maps are difficult to generate dynamic explainable recommendation logic. To address the aforementioned issues, this paper proposed a framework of explainable learning path recommendation based on dynamic attention reinforcement learning (ELPR-DARL). Firstly, a heterogeneous collaborative knowledge graph was constructed, integrating course text, visual content, and knowledge dependencies to enhance cross-modal semantic alignment capabilities. Secondly, a dynamic attention aggregation mechanism for adjacent nodes was designed, which adjusts the weights of entity relationships through a bias correction strategy, and a bidirectional interaction aggregator was utilized to fuse multi-level neighborhood features, enhancing the fine-grained expression ability of knowledge reasoning. Finally, a knowledge graph-aware reinforcement learning strategy was proposed, which explicitly modelled the association between user behavior and knowledge topology based on path connectivity reward functions, generating explainable paths that include global rewards and local attention weights. Experiments based on the MOOC dataset show that this method achieves 22.85%, 33.81%, 52.01%, and 6.34% in NDCG, Recall, HR, and precision metrics, respectively, which is 2.88%, 3.55%, 2.42%, and 3.26% higher than the suboptimal model. User research shows that 80.36% of learners believe that path explanation significantly improves recommendation transparency. This study verifies that the collaborative optimization of a dynamic attention mechanism and reinforcement learning can effectively balance recommendation accuracy and explainability.

Key words: collaborative knowledge graph, learning path recommendation, explainable recommendation, dynamic attention mechanism, reinforcement learning, recommendation system

中图分类号:

TP391

张晓明, 冯泽嘉, 王会勇, 张晓静. 基于动态注意力强化学习的可解释学习路径推荐[J]. 应用科学学报, 2026, 44(1): 110-133.

ZHANG Xiaoming, FENG Zejia, WANG Huiyong, ZHANG Xiaojing. Explainable Learning Path Recommendation Based on Dynamic Attention Reinforcement Learning[J]. Journal of Applied Sciences, 2026, 44(1): 110-133.

[1] Aljunid M F, D H M, Hooshmand M K, et al. A collaborative filtering recommender systems: Survey [J]. Neurocomputing, 2025, 617: 128718.
[2] Zheng X Y, Ni Z, Zhong X N, et al. Kernelized deep learning for matrix factorization recommendation system using explicit and implicit information [J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(1): 1205-1216.
[3] Kizilcec R F. How much information: effects of transparency on trust in an algorithmic interface [C]//2016 CHI Conference on Human Factors in Computing Systems, 2016: 2390- 2395.
[4] Zhang Y F, Lai G K, Zhang M, et al. Explicit factor models for explainable recommendation based on phrase-level sentiment analysis [C]//37th International ACM SIGIR Conference on Research & Development in Information Retrieval, 2014: 83-92.
[5] Wu Y, Ester M. FLAME: a probabilistic model combining aspect based opinion mining and collaborative filtering [C]//18th ACM International Conference on Web Search and Data Mining, 2015: 199-208.
[6] Liu S X, Fan C J, Cheng K W, et al. Inductive meta-path learning for schema-complex heterogeneous information networks [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(12): 10196-10209.
[7] Wang X T, Chen Y R, Yang J, et al. A reinforcement learning framework for explainable recommendation [C]//2018 IEEE International Conference on Data Mining (ICDM), 2018: 587- 596.
[8] Chang S, Harper F M, Terveen L G. Crowd-based personalized natural language explanations for recommendations [C]//10th ACM Conference on Recommender Systems, 2016: 175- 182.
[9] Markchom T, Liang H Z, Ferryman J. Review of explainable graph-based recommender systems [J]. ACM Computing Surveys, 2026, 58(6): 1-35.
[10] Tian R Y, Cai J J, Li C Z, et al. Self-supervised pre-training model based on multi-view for MOOC recommendation [J]. Expert Systems with Applications, 2024, 252: 124143.
[11] Li C H, Luo Y. Integrating knowledge graph reasoning and reinforcement learning for explainable MOOC recommendations [J]. IEEE Access, 2025, 13: 183722-183733.
[12] Lyu Z Y, Wu Y, Lai J J, et al. Knowledge enhanced graph neural networks for explainable recommendation [J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(5): 4954- 4968.
[13] Guan Q L, Cheng X H, Xiao F, et al. Explainable exercise recommendation with knowledge graph [J]. Neural Networks, 2025, 183: 106954.
[14] Balloccu G, Boratto L, Fenu G, et al. Reinforcement recommendation reasoning through knowledge graphs for explanation path quality [J]. Knowledge-Based Systems, 2023, 260: 110098.
[15] Xian Y K, Fu Z H, Muthukrishnan S, et al. Reinforcement knowledge graph reasoning for explainable recommendation [C]//42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, 2019: 285-294.
[16] Frej J, Shah N, Knezevic M, et al. Finding paths for explainable MOOC recommendation: a learner perspective [C]//14th Learning Analytics and Knowledge Conference, 2024: 426-437.
[17] Guan Q L, Xiao F, Cheng X H, et al. KG4Ex: an explainable knowledge graph-based approach for exercise recommendation [C]//32nd ACM International Conference on Information and Knowledge Management, 2023: 597-607.
[18] Lin Y G, Zhang W, Lin F, et al. Knowledge-aware reasoning with self-supervised reinforcement learning for explainable recommendation in MOOCs [J]. Neural Computing and Applications, 2024, 36(8): 4115-4132.
[19] Afreen N, Balloccu G, Boratto L, et al. Learner-centered ontology for explainable educational recommendation [C]//Adjunct Proceedings of the 32nd ACM Conference on User Modeling, Adaptation and Personalization, 2024: 567-575.
[20] Ai Q Y, Azizi V, Chen X, et al. Learning heterogeneous knowledge base embeddings for explainable recommendation [J]. Algorithms, 2018, 11(9): 137.
[21] Zhang Y F, Zhang Y W, Cao X L, et al. Interpretable MOOC course recommendations based on reinforcement knowledge graph [C]//20242nd International Conference on Big Data and Privacy Computing (BDPC), 2024: 11-15.
[22] Zhang J, Hao B W, Chen B, et al. Hierarchical reinforcement learning for course recommendation in MOOCs [J]. AAAI Conference on Artificial Intelligence, 2019, 33(1): 435-442.
[23] Polyzou A, Nikolakopoulos A N, Karypis G. Scholars walk: a Markov chain framework for course recommendation [C]//12th International Conference on Educational Data Mining, 2019: 396-401.
[24] Tiwary N, Mohd Noah S A, Fauzi F, et al. A review of explainable recommender systems utilizing knowledge graphs and reinforcement learning [J]. IEEE Access, 2024, 12: 91999-92019.
[25] Khalid A, Lundqvist K, Yates A, et al. Online learning path recommender system for MOOCs [C]//2023 IEEE Global Engineering Education Conference (EDUCON), 2023: 1-10.
[26] Jiang B, Li X Y, Yang S H, et al. Data-driven personalized learning path planning based on cognitive diagnostic assessments in MOOCs [J]. Applied Sciences, 2022, 12(8): 3982.
[27] Joseph L, Abraham S, Mani B P, et al. Exploring the effectiveness of learning path recommendation based on felder-silverman learning style model: a learning analytics intervention approach [J]. Journal of Educational Computing Research, 2022, 60(6): 1464-1489.
[28] Niknam M, Thulasiraman P. LPR: a bio-inspired intelligent learning path recommendation system based on meaningful learning theory [J]. Education and Information Technologies, 2020, 25(5): 3797-3819.
[29] Van Houdt G, Mosquera C, Nápoles G. A review on the long short-term memory model [J]. Artificial Intelligence Review, 2020, 53(8): 5929-5955.
[30] Yang Z J, Wen J H, Abdulkadir A, et al. Gene-SGAN: discovering disease subtypes with imaging and genetic signatures via multi-view weakly-supervised deep clustering [J]. Nature Communications, 2024, 15: 354.
[31] Cheng Y J. A learning path recommendation method for knowledge graph of professional courses [C]//IEEE 22nd International Conference on Software Quality, Reliability, and Security Companion (QRS-C), 2022: 469-476.
[32] Yin H, Sun Z Y, Sun Y C, et al. Automatic learning path recommendation for open source projects using deep learning on knowledge graphs [C]//2021 IEEE 45th Annual Computers, Software, and Applications Conference (COMPSAC), 2021: 824-833.
[33] Bordes A, Usunier N, García-Durán A. Translating embeddings for modeling multirelational data [C]//Neural Information Processing Systems, 2013: 2787-2795.
[34] Wei H, Liu C, Chen J, et al. General OCR theory: towards OCR-2.0 via a unified end-to-end model [DB/OL]. (2024-09-03) [2025-08-11]. https://arxiv.org/abs/2409.01704.
[35] Devlin J, Chang M W, Lee K, et al. BERT: pre-training of deep bidirectional transformers for language understanding [C]//North American Chapter of the Association for Computational Linguistics, 2019: 4171–4186.
[36] Xu W J, Li H, Wang M H. Multi-behavior guided temporal graph attention network for recommendation [M]//Advances in Knowledge Discovery and Data Mining. Cham: Springer Nature Switzerland, 2023.
[37] Maas Al, Hannun A Y, Ng A Y. Rectifier nonlinearities improve neural network acoustic models [C]//30th International Conference on Machine Learning (ICML’13)—Workshop on Deep Learning for Audio, Speech and Language Processing. JMLR W&CP, 2013, 28: 3.
[38] Yu J F, Luo G, Xiao T, et al. MOOCCube: a large-scale data repository for NLP applications in MOOCs [C]//58th Annual Meeting of the Association for Computational Linguistics, 2020: 3135-3142.
[39] He X N, Liao L Z, Zhang H W, et al. Neural collaborative filtering [C]//26th International Conference on World Wide Web, 2017: 173-182.
[40] Zhang P Y, Yan Y C, Zhang X, et al. TransGNN: harnessing the collaborative power of transformers and graph neural networks for recommender systems [C]//International ACM SIGIR Conference on Research and Development in Information Retrieval, 2024: 1285-1295.

基于动态注意力强化学习的可解释学习路径推荐

Explainable Learning Path Recommendation Based on Dynamic Attention Reinforcement Learning

PDF

可视化

摘要/Abstract

引用本文

使用本文

相关文章 15

编辑推荐

Metrics

本文评价

[1]	郭艳, 王浩然, 侯嵩松, 段旭良, 穆炯. 超图神经网络驱动的跨视角群组推荐算法[J]. 应用科学学报, 2026, 44(3): 486-502.
[2]	伊华伟, 宋仕玺, 王艳飞, 白思怡. 融合图神经网络和深度图聚类的联邦推荐算法[J]. 应用科学学报, 2026, 44(1): 83-96.
[3]	卢菁, 葛聪. 结合通用轨迹图和多偏好的兴趣点推荐方法[J]. 应用科学学报, 2025, 43(5): 771-784.
[4]	朱耀辉, 王涛, 彭振春, 刘含. 无人机辅助NOMA通信系统中的3D轨迹优化与资源分配[J]. 应用科学学报, 2025, 43(2): 208-221.
[5]	刘珈宁, 张思佳, 张正龙, 王祎涵, 安宗诗. 基于多跳机制的扩散图谱推荐模型[J]. 应用科学学报, 2025, 43(1): 123-136.
[6]	艾均, 洪星琦. 融合标题情感和话题特征的新闻推荐算法[J]. 应用科学学报, 2024, 42(5): 810-822.
[7]	金彦亮, 范宝荣, 高塬, 汪小勇, 顾晨杰. 基于元学习和强化学习的自动驾驶算法[J]. 应用科学学报, 2024, 42(5): 795-809.
[8]	苏湛, 杨昊川, 艾均. 基于模糊偏好标签向量的推荐算法[J]. 应用科学学报, 2024, 42(3): 525-539.
[9]	刘星彤, 郑红, 黄建华. 一种改进近端优化的多目标流QoS调度策略[J]. 应用科学学报, 2024, 42(3): 499-512.
[10]	王鑫, 仲伟志, 王俊智, 肖丽君, 朱秋明. 基于深度强化学习的无人机路径规划与无线电测绘[J]. 应用科学学报, 2024, 42(2): 200-210.
[11]	葛星, 秦丽, 沙瀛. 基于投影奖励机制的多机器人协同编队与避障[J]. 应用科学学报, 2024, 42(1): 174-188.
[12]	张华南, 李石君, 金红. 无线传感器网络强化学习增强路由研究[J]. 应用科学学报, 2024, 42(1): 83-93.
[13]	苏湛, 陈学谦, 艾均, 黄忠. 基于用户相似性选择及标签距离的推荐算法[J]. 应用科学学报, 2023, 41(6): 940-957.
[14]	彭凯, 刘培琛, 许小龙, 周星宇. 面向智慧城市的多依赖任务计算迁移研究[J]. 应用科学学报, 2023, 41(3): 391-404.
[15]	季德强, 王海荣, 车淼, 王嘉鑫. KNN-GWD推荐模型及其应用[J]. 应用科学学报, 2022, 40(1): 145-154.