원문정보
초록
영어
The user-based collaborative filtering algorithm has been widely used in various kinds of personalized recommendation systems. But it has a serious shortcoming: with the increasing number of the users and commodities, its calculation work grows rapidly. To address the problem of vast time consumption by big dataset, we utilize MapReduce programming idea to do parallelized transformation of the algorithm; finally deploy it to be run in Hadoop cloud computing platform. Experiments have revealed that if computing data is reasonably distributed and the data volume is big, then the algorithm performance of the algorithm can realize favorable linearly speeding effect.
목차
1. Introduction
2. Collaborative Filtering Algorithm
3. Description of Collaborative Filtering System
3.1. Hypothesis and Objective
3.2. Specific Process of Collaborative Filtering Algorithm
3.3. Problems Facing the Traditional Collaborative Filtering Algorithm and Solutions
4. MapReduce Parallelization of Traditional Collaborative Filtering Algorithm
4.1 Data division
4.2. Map Stage
4.3. Reduce Stage
5. Experimental Analysis and Results
6. Conclusion
References