关键词: 关键词: Slope One, 聚类, Spark, 时间权重, 项目属性

Abstract: Abstract: Focused on the problem that without putting the item attribute and time factor into account the similarity between items in traditional Slope One algorithm, as well as the recommended bottleneck that slow processing and high complexity under the big data background, a weighted Slope One algorithm based on clustering and Spark framework is put forward. Firstly, the rating similarity was included into the time weight and comprehensive similarity was computed by introducing item attribute. And then the set of nearest neighbor was generated through combining with the Canopy-K-means algorithm; Finally, the data is partitioned and iterated to realize parallelization by using of Spark framework. Experimental results show that the improved algorithm based on Spark framework can significantly increase in accuracy of predicted rating compared with traditional Slope One algorithm and Slope One based on user similarity, averagely improve 3.5-5 times in running efficiency compared with Hadoop platform, and is more suitable for recommendation of the large data.

Key words: Keywords: Slope One, clustering, spark, time weight, item attribute