基于社交关系与矩阵补全的协同过滤的推荐算法研究

发布时间：2018-07-18 08:17

【摘要】：近年来,计算机网络通信等技术日益精湛,人们已经迈进了一个全新的大数据时代,大数据时代的到来进一步增加了数据膨胀的程度,传统的信息检索系统和推荐系统已经无法满足大数据环境下的检索需求。推荐系统的诞生从某些方面解决了搜索引擎存在的搜索结果不准确等问题。推荐算法是一个推荐系统的核心,协同过滤推荐算法是推荐系统领域最经典的算法之一。但是在这个全新的大数据的背景下,协同过滤推荐算统也存在着一定的问题,它无法很好地解决评分矩阵稀疏和冷启动问题。本文针对这两个问题做了一定的改进,将社交关系的思想加入到协同过滤的推荐过程中,并对矩阵补全的方式进行了一定的改进,主要工作如下:一方面,将社交关系思想加入到相似度计算过程中,得到了基于社交关系的协同过滤推荐算法。通过社交关系数据能够得到每个用户的好友集合,根据目标用户朋友们喜欢的项目或内容为目标用户进行推荐,该算法可以较好地解决基于用户的冷启动,能够提高新用户的推荐满意度,本文通过实验对此进行了验证。另一方面,条件性地筛选矩阵补全的位置,得到基于社交关系和条件性补全的协同过滤推荐算法。该算法在融入社交关系思想的基础上对矩阵补全方式进一步改进,选择满足一定条件的项进行补全,使得补全后的矩阵更准确,减少了数据冗余,该方法能够很好地解决数据稀疏问题,提高算法效率并且使推荐准确率得以提高。本文通过具体的实验对改进的算法进行了定量的分析和对比,实验结果证明改进的算法具有更高的推荐准确率和推荐效率,并且具有较好的MAE、MRSE值。
[Abstract]:In recent years, computer network communication and other technologies have become increasingly sophisticated, people have entered a new era of big data, the arrival of the big data era has further increased the degree of data expansion. Traditional information retrieval systems and recommendation systems can no longer meet the retrieval requirements under the big data environment. The birth of recommendation system solves the problem of inaccuracy of search results in some aspects. Recommendation algorithm is the core of a recommendation system, and collaborative filtering recommendation algorithm is one of the most classical algorithms in the field of recommendation system. However, under the background of this new big data, there are some problems in collaborative filtering recommendation system, which can not solve the problem of sparse score matrix and cold start. In this paper, some improvements have been made to these two problems, the idea of social relations has been added to the recommendation process of collaborative filtering, and the method of matrix complement has been improved. The main work is as follows: on the one hand, The idea of social relationship is added to the similarity calculation process, and a collaborative filtering recommendation algorithm based on social relationship is obtained. Through the social relationship data, each user's friend set can be obtained, and the target user can be recommended according to the items or content that the target user's friends like. The algorithm can solve the cold start based on the user better. It can improve the new user's recommendation satisfaction, this article has carried on the verification through the experiment. On the other hand, a collaborative filtering recommendation algorithm based on social relationship and conditional complement is obtained by selecting the position of matrix complement conditionally. The algorithm further improves the method of matrix complement based on the idea of social relations, and selects the items that meet certain conditions to complement the matrix, which makes the matrix more accurate and reduces the data redundancy. This method can solve the problem of data sparsity, improve the efficiency of the algorithm and improve the accuracy of recommendation. In this paper, quantitative analysis and comparison of the improved algorithm are carried out through specific experiments. The experimental results show that the improved algorithm has higher recommendation accuracy and efficiency, and better MAE / MRSE value.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【相似文献】