摘要
针对Slope One算法存在预测精度依赖于用户对待预测项目的评分数量的缺陷,提出了一种基于项目属性相似度和Map Reduce并行化的Slope One算法.首先计算项目间的属性相似度,并将其与Slope One算法相融合以提高预测精度,然后在Hadoop平台上对改进算法基于Map Reduce进行并行化实现.在Movie Lens数据集上的实验结果表明,相对于Slope One算法和加权Slope One算法,本文提出的改进Slope One算法具有更高的预测精度,并更适用于大规模数据集.
Directing at the Slope One algorithm’s drawback that the predicted precision relies on the number of users’ratings to the predicted item, this paper presents an improved Slope One algorithm based on the item’s attribute similarity and MapReduce in parallel. In this proposed algorithm. We firstly compute the attribute similar-ity between the items, and combine it with the Slope One algorithm to improve the prediction precision, and next, implement the parallel algorithm based on MapReduce over the Hadoop platform. Experimental results on the MovieLens data set show that the improved Slope One algorithm is of higher predicted precision, and is more suit-able for large-scale data set, compared with the Slope One algorithm and the weighted Slope One algorithm.
出处
《空军预警学院学报》
2015年第1期54-58,67,共6页
Journal of Air Force Early Warning Academy