推荐系统主要由两个方法组成,即基于内容属性相似的推荐和基于协同过滤算法的推荐,这两种方法可以提供有意义的推荐.其中基于内容属性相似的推荐只是单纯地依赖物品之间的属性相似来构建推荐关系;而协同过滤算法推荐作为推荐系统领域的经典,它不会去研究物品的本身属性,正如名字描述的一样,严重依靠于用户的行为及其周边用户的协同行为.文中使用了一种改进的混合方法,充分考虑和利用了协同过滤算法和内容属性过滤的优点.讨论的算法与该领域以前的方法是不同的,因为它包括一个新颖的方法来找到两个事物之间的相似内容.包含了一个分析用以证明这个新的方法,并且阐述了它是怎样提供实用性的推荐的.与其他两种常用的方法进行比较,即纯协同过滤(Pure CF)和奇异值分解(SVD),结果表明,该方法经过现有的用户和目标数据的测试,产生了有所改进的结果.%The recommendation system is mainly composed of two methods, namely, the recommendation based on the similarity of content attributes and based on a collaborative filtering algorithm, which can help in providing meaningful recommendations. Among them, the former is simply relying on attribute similarity between items to build recommendation relationships, while the latter is recommended as a classic of the field of recommendation systems, not studying the attributes of items itself, just as its name described, relying heavily on the users' behavior and the collaborative behavior of surrounding users. In this article, we use an improved hybrid approach that fully considers and exploits the advantages of collaborative filtering algorithms and content attribute filtering. The algorithm discussed in this article is different from the previous method in this field because it includes a novel method to find similar content between two items. An analysis is contained to demonstrate this new method, discussing how it provides practical recommendations. Compared with the other two commonly used methods, Pure CF and SVD, it shows that the method in this paper results in improved results by passing the test of existing users and target data.
展开▼
