高校图书推荐系统算法与模型的研究

发布时间：2018-09-07 16:55

【摘要】：21世纪以来,全世界科技水平不断提高,信息呈现爆炸式增长,人们从寻找信息的模式变成了寻找有用信息的模式。从海量信息中寻找到对自己有用信息的手段有很多,推荐系统是其中最重要也是最广泛使用的手段之一。各大网商通过使用推荐系统对用户进行个性化的推荐,都取得了不错的成果,这也为推荐系统在高等院校图书推荐领域的应用提供了可能性。推荐系统中的算法有很多,其中最经典应用最广泛的是协同过滤算法。本文对基于用户和基于项目的协同过滤算法进行了深入研究,针对高校图书推荐的特殊性,如借阅数据而无法直接使用,相似度矩阵太过稀疏而无法产生推荐等问题,改进了这两种算法。但是这两个算法在高校图书推荐领域都有着各自的优劣势,通过对两者的结合,提出了混合推荐系统模型。最后通过实验对比了混合推荐算法与单一推荐算法的各项评价指标,为应用于高校图书推荐提供了理论支撑。本研究的主要工作有以下五个部分:第一部分,深入研究了推荐系统的原理以及一些经典的推荐算法,并对推荐算法应用在高校图书推荐领域的可行性进行了分析,然后构建了基于读者(RCF)和基于图书(BCF)的协同过滤算法模型。第二部分,由于高校图书推荐不同于电影推荐或者商品推荐,它不包含用户对物品的评分,针对这一特点,通过对图书借阅记录的处理,提出一种量化评分模型,将读者对图书的偏好定量化,在此基础上,构建了读者-图书的评分矩阵。第三部分,针对读者-图书评分矩阵过于稀疏的特点,将中文图书分类法与图书借阅记录相结合,构建了读者-图书类别评分矩阵,然后在此基础上改进了基于读者和基于图书的协同过滤算法。第四部分,通过对改进后的RCF和BCF结合,构建了混合推荐系统(HCF)模型,然后进行了实验验证,并评估了三种算法模型。第五部分,根据模型的结果,对高校图书推荐系统的应用提出了自己的建议。
[Abstract]:Since the 21st century, the level of science and technology all over the world has been improved, and the information has been increasing explosively. People have changed from the mode of searching for information to the mode of seeking useful information. There are many ways to find useful information from mass information, and recommendation system is one of the most important and widely used methods. By using the recommendation system, all the network merchants have achieved good results, which provides the possibility for the application of the recommendation system in the field of book recommendation in colleges and universities. There are many algorithms in recommendation system, among which the most classical one is collaborative filtering algorithm. In this paper, the collaborative filtering algorithm based on user and item is studied in depth. Aiming at the particularity of book recommendation in colleges and universities, such as borrowing data and not using it directly, the similarity matrix is too sparse to produce recommendation and so on. These two algorithms are improved. However, these two algorithms have their own advantages and disadvantages in the field of book recommendation in colleges and universities. Through the combination of the two algorithms, a hybrid recommendation system model is proposed. Finally, the evaluation indexes of mixed recommendation algorithm and single recommendation algorithm are compared through experiments, which provides theoretical support for the application of book recommendation in colleges and universities. The main work of this study is as follows: in the first part, the principle of recommendation system and some classical recommendation algorithms are studied in depth, and the feasibility of applying recommendation algorithm in the field of book recommendation in colleges and universities is analyzed. Then a collaborative filtering algorithm model based on reader (RCF) and book (BCF) is constructed. The second part, because the university book recommendation is different from the movie recommendation or the commodity recommendation, it does not contain the user to the article score, in view of this characteristic, through the processing to the book loan record, proposed one kind of quantification score model. On the basis of quantifying readers' preference for books, a reader-book scoring matrix is constructed. In the third part, aiming at the characteristic that the reader-book scoring matrix is too sparse, this paper combines the Chinese book classification method with the book borrowing record, and constructs the reader-book classification scoring matrix. Then the collaborative filtering algorithm based on readers and books is improved. In the fourth part, through the combination of improved RCF and BCF, the (HCF) model of hybrid recommendation system is constructed, and then the experimental verification is carried out, and three kinds of algorithm models are evaluated. In the fifth part, according to the results of the model, the author puts forward some suggestions on the application of the book recommendation system in colleges and universities.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】