AN IMPROVED HYBRID RECOMMENDATION ALGORITHM STUDY AND DESIGN ON MAHOUT FRAMEWORK

Dai Fei, Xiaohui Cheng

ABSTRACT: Collaborative filtering algorithm is one of the most widely used algorithms in the recommendation system. However, each collaborative filtering algorithm has its advantages and disadvantages. When applied to recommendation systems alone, it leads to low efficiency and low accuracy. Based on the research of existent recommendation algorithms, we improved a hybrid recommendation algorithm using Mahout Framework. This improved algorithm consists of third parts, first part: recommended results were got by using a user-based collaborative filtering algorithm to compute; second part: recommended results were got by using the improved item-based collaborative filtering algorithm to compute; third part: the two recommendation results were tested by the screening mechanism, then output the best recommended results. This paper’s innovation point is that we use singular value decomposition and principal component analysis algorithm to optimize dimension reduction on item-based collaborative filtering algorithm and get better effect of the recommend system. Compared with the originally ex perimental results, recommendation system of the improved hybrid recommendation algorithm can have more precise outcome and high efficiency.

Keywords: Collaborative filtering algorithm, Mahout, singular value decomposition, hybrid recommendation algorithm