基于偏好的教育推荐系统设计与实现

发布时间：2018-08-28 12:15

【摘要】：互联网使用的增长直接催生了信息过载。大量可用的话题出现在网上使得相关信息源的搜索变得困难和耗时。在线教育是传统教育模式的一种变革,用户可以根据自己的学习节奏在虚拟环境中学习,但由于当前的搜索引擎和教育网络很难在短时间内为知识的搜寻者提供他们真正需要的资源,资源获取极其耗时成了一个大问题。另外,目前的在线教育交互性差,偏向于快速适应和趣味性,也使得用户的忠诚度非常低。当前环境下,很多研究人员通过引入社会网络给用户推荐与他们的选择和历史行为相一致的资源,其中利用到协同过滤,基于内容的过滤以及混合推荐算法。早期的方法强调为特定用户计算基于内容相关度的重要性,系统为用户提供相关度得分最高的资源给用户。评分计算方法围绕内容的多个属性计算评分,不同的内容相关度评分计算结果差异较大。例如,社会网络中认为用户的亲密度和历史浏览行为模式在计算评分中更为重要,而在一般的推荐系统中,更侧重于有共同属性的相似用户,如位置、教育程度、隐私设置等。推荐系统中一个最主要的问题是冷启动,数据缺失或者稀疏数据成为推荐系统的瓶颈。本文在现有的社会网络的基础上提出了一种协作、共享、排名和推荐的在线教育系统,根据用户的个人偏好进行相应的推荐。系统首先通过用户注册模块获取用户的相关信息,具有相同偏好的相似用户被分到同一个组中。然后基于相似用户群组,计算出资源的相关性得分,把得分最高的资源推荐给用户。针对推荐系统的冷启动问题,也就是说,如果用户没有设置相应的偏好信息,系统则给用户推荐最近热门的资源并从用户的反馈行为中学习。在这样系统中,用户接口设计是极其重要的部分,可以说是整个系统的重心,用户体验良好与否直接影响整个系统的成败。系统的质量取决于用户界面的易使用、易理解、易交互,并且用户界面根据用户需求是完全可以定制的。通过这样的方式设计系统,可以减少每个资源遍历的负载,提高用户的效率,并得出最合适的结果。综合考虑用户当前的搜索内容和用户偏好使得资源的搜索更加高效,这是当前在线学习模式的变革,也是未来在线教育的一个关键技术点。图35副,表5个,参考文献60篇。
[Abstract]:The growth of Internet use directly spawned information overload. The availability of a large number of topics on the Internet makes it difficult and time-consuming to search for relevant sources of information. Online education is a kind of transformation of traditional education mode. Users can learn in virtual environment according to their own learning rhythm. However, because the current search engines and educational networks are difficult to provide the knowledge searchers with the resources they really need in a short period of time, obtaining resources becomes a big problem. In addition, current online education is less interactive, more adaptive and more interesting, making users' loyalty very low. In the current environment, many researchers introduce social networks to recommend resources consistent with their choices and historical behaviors, which make use of collaborative filtering, content-based filtering and hybrid recommendation algorithms. Early methods emphasized the importance of computing content-based relevance for specific users, and the system provided users with resources with the highest relevance score. The method of score calculation is based on the multiple attributes of content, and the results of different content correlation scores are quite different. For example, in social networks, user affinity and historical browsing behavior patterns are considered to be more important in the calculation of ratings, while in general recommendation systems, similar users with common attributes, such as location, educational level, privacy settings, etc. One of the most important problems in recommendation system is cold start. Missing or sparse data becomes the bottleneck of recommendation system. Based on the existing social networks, this paper proposes an online education system based on collaboration, sharing, ranking and recommendation, which can be recommended according to users' personal preferences. The system first obtains the relevant information of the user through the user registration module, and similar users with the same preference are divided into the same group. Then, based on the similar user groups, the correlation score of resources is calculated, and the resources with the highest score are recommended to the users. For the cold start problem of recommendation system, that is, if the user does not set the corresponding preference information, the system recommends the most popular resources to the user and learns from the feedback behavior of the user. In such a system, user interface design is an extremely important part, can be said to be the focus of the whole system, the user experience is good or not directly affect the success or failure of the whole system. The quality of the system depends on the user interface is easy to use, easy to understand, easy to interact, and the user interface can be completely customized according to user needs. By designing the system in this way, the load of each resource traversal can be reduced, the efficiency of the user can be improved, and the most suitable result can be obtained. The comprehensive consideration of users' current search content and user preferences makes the search of resources more efficient, which is the change of current online learning mode and a key technology point of online education in the future. There are 35 pairs of figs, 5 tables and 60 references.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3

【共引文献】