【作者】 赵鑫；

【导师】 王龙；

【作者基本信息】 辽宁大学 ， 软件工程（专业学位）， 2023， 硕士

【摘要】 近年来,随着网络技术的普及以及教育行业的迅速发展,越来越多的教学内容逐渐通过网络教学的方式进行传递。在大数据背景下,网络化教学的趋势逐渐显现。如何在高效和稳定的情况下,满足学生和用户的需求,尤其是在信息过载的情况下,如何为学生精确推荐相关教育资源,成为网络在线化教学需要解决的一个问题。并且在推荐领域中,如何有效地确定相似度是协同过滤技术的核心问题,因为相似度不仅决定了对邻居用户的选择,而且对推荐质量有着决定性的影响。同时如何利用优质的优化算法去筛选候选项成为一种重要策略。本文基于以上理论进行了如下工作:首先,本文从理论学习的角度出发,对移动端学习基础、小程序开发以及推荐算法相关的国内外研究论文进行分析。接着结合实际情况,分析了当前受欢迎的教育APP和小程序所采用的推荐算法。在此基础上,进行了在线教育系统的需求分析工作,并设计了系统的基本架构,确立了本系统需要实现的功能。其次,为了更好地为用户推荐相关教育资源,本文基于两个信号源差异的α-散度相似度算法,针对该算法进行改进。首先构建用户评分矩阵,计算出α-散度相似度,随后基于α-散度本身存在的一些特性,提出一个公式将其与课程标签相似度结合,提出了一种将α-散度相似度与课程标签相似度结合的方案,利用该相似度推荐出一次候选项,从而保证相似度的精确度。同时针对优化方面,从用户维度进行入手,通过训练随机森林模型对一次候选项进行二次优化操作,以便于通过用户的个人特征为用户更好地推荐教育资源,同时在随机森林模型内部,提出一种基于边界值及二次训练的随机森林加权模型,增加随机森林分类的准确率,提高系统的推荐精度。最后,将本文提出的推荐算法应用到了在线教育系统当中,设计并实现了在线教育系统。根据算法的实际需求,实现了推荐系统模块以及其他相关功能,客户端主要基于微信小程序进行开发,采用小程序的方式主要增强了用户的便捷性,可以使用户能够随时随地学习到自己需要的资源。经过测试证明,本系统基本功能完善,系统运行稳定,能够满足用户对学习资源的偏好需求。更多还原

【Abstract】 In recent years,with the popularization of network technology and the rapid development of education industry,more and more teaching content is gradually transmitted through online teaching.In the context of big data,the trend of network teaching gradually appears.How to meet the needs of students and users in high efficiency and stability,especially in the case of information overload,how to accurately recommend relevant educational resources for students,has become a problem that needs to be solved in online teaching.And in the field of recommendation,how to effectively determine the similarity is the core issue of collaborative filtering technology,because the similarity not only determines the selection of neighbor users,but also has a decisive impact on the quality of recommendation.At the same time,how to use high quality optimization algorithm to select candidates becomes an important strategy.Based on the above theories,this paper carries out the following work:Firstly,from the perspective of theoretical learning,this paper analyzes domestic and foreign research papers on the basis of mobile terminal learning,small program development and recommendation algorithm.Then combined with the actual situation,the recommendation algorithm used by the popular educational apps and small programs is analyzed.On this basis,the requirements of the online education system are analyzed,and the basic structure of the system is designed,and the functions of the system need to be realized are established.Secondly,in order to better recommend relevant educational resources to users,this article is based on the differences between two signal sources α-divergence Improve the divergence similarity algorithm.Firstly,construct a user rating matrix and calculate α-divergence similarity,subsequently based on α-divergence,A formula is proposed to combine the inherent characteristics of divergence with the similarity of course labels,and a method is proposed to α-divergence,A scheme combining divergence similarity with course label similarity is proposed to recommend candidate options using this similarity,thereby ensuring the accuracy of similarity.At the same time,in terms of optimization,starting from the user dimension,the random forest model is trained to conduct secondary optimization operations on primary candidates,so as to better recommend educational resources for users through the user’s personal characteristics.At the same time,within the random forest model,a random forest weighted model based on boundary values and secondary training is proposed to increase the accuracy of random forest classification and improve the recommendation accuracy of the system.Finally,the recommendation algorithm proposed in this article was applied to an online education system,and an online education system was designed and implemented.According to the actual requirements of the algorithm,the recommendation system module and other related functions have been implemented.The client is mainly developed based on We Chat mini programs,which mainly enhance user convenience and enable users to learn the resources they need anytime and anywhere.The test proves that the basic function of the system is perfect,the system runs stably,and can meet the user’s preference for learning resources.更多还原