农业科技文献推荐算法及评估
首发时间:2018-12-10
摘要:推荐系统在互联网普及与发展带来了信息超载问题的背景下产生的,它主要是根据用户的兴趣和特点,向用户推荐其最可能感兴趣的信息或者商品,从而解决信息超载的问题。文献在农业科学研究中起着重要作用,随着文献数量的增长,为使用户快速便捷地找到相关文献,文献推荐系统成为了农业科学院知识系统重要的组成部分。传统的推荐方法,包括基于内容的推荐、协同过滤和混合推荐。本文采用的是基于内容和协同过滤相结合的混合推荐算法。首先对文献数据进行分词、权重计算,然后用余弦相似度的算法对文献进行相似度计算,最后用User-Based、Item-Based和SVD三种推荐算法进行文献推荐,得到推荐结果,并用准确率和召回率两种测试指标对三种推荐算法进行评估。实验表名SVD算法与另外两种算法相比准确率和召回率都较高。
For information in English, please click here
Recommendation method and evaluation of agricultural science and Technology Literature
Abstract:Recommendation system comes into being under the background of information overload that brought about by the popularization and development of the Internet. It mainly refers information or commodities that users are most likely interested in according to their interests and characteristics, so as to solve the problem of information overload. Literature plays an important role in agricultueal scientific research. With the growth in the number of literature, fthe literature recommendation system has become an important part of the knowledge system of the Academy of Agricultural Sciences in order to enable users to find relevant documents quickly and conveniently. Traditional recommendation methods, including content-based recommendation, collaborative filtering and hybrid recommendation. This paper adopts a hybrid recommendation algorithm based on content and collaborative filtering. Firstly, we compute the word segmentation and weight of the literature data, and then use cosine similarity algorithm to calculate the similarity of the literature. Finally, we use User-Based, Item-Based and SVD to recommend the literature, and get the recommendation results. Finally, we evaluate the three recommendation algorithms with two test indicators: accuracy and recall. Experiments show that SVD algorithm has higher accuracy and recall rate than the other two algorithms.
Keywords: Hybrid recommendation cosine similarity TF-IDF algorithm
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
农业科技文献推荐算法及评估
评论
全部评论0/1000