摘要
讨论基于Mahout的推荐系统开发过程,以我校学生对学校食堂各就餐窗口进行评分为例,通过建立实验系统,利用Mahout推荐系统引擎提供的API分析实验数据。该实验样例完整地阐述创建自己的推荐引擎构造器的过程。本实验采用常用的协同过滤推荐算法,协同过滤推荐算法主要包括基于用户的协同过滤,基于物品的协同过滤以及SlopeOne推荐算法。搭建自己的开发环境,基于相似度算法模型和推荐算法,使用7种不同组合进行对比实验。使用查准率和召回率两个指标对7种算法组合进行评估。采用欧氏距离用户相似度,基于用户和物品的推荐算法,并且采用有评分和无评分的方法对推荐结果进行比较,由实验得知,基于Mahout的推荐系统能快速高效地给学生推荐相似的就餐窗口。
Constructs an experimental system based on the deep discussion on development process of Mahout recommended system and empirical researches on scoring all dining windows by students. Furthermore, conducts a deep analysis on experimental data through API provided by Mahout recommended system engine. This experimental sample completely narrates the process to create its own recommended engine constructor. The commonly-used collaborative filtering recommendation algorithm is adopted by this experiment which includes collabora-tive filtering of users, collaborative filtering of articles and recommendation algorithm of SlopeOne. Under construction of its own develop-ment environment, adopts seven different combinations to conduct contrast experiment based on similarity algorithm model and recom-mended algorithm. Precision ratio and recall rate are applied to evaluate the seven algorithm combinations. Compares these recommended results through Euclidean distance similarity of users, collaborative filtering of users, collaborative filtering of articles, scoring method and non-scoring method. From the experimental results, it can be concluded that recommended system of Mahout can efficiently and rapidly recommend similar dining windows to students.
