摘要
针对传统推荐算法在运算速度及稳定性不足等问题提出了基于矩阵模型的创新算法.通过对手机社区用户图书近一年的下载数据进行分析,依次测试每个月不同数据量下新旧算法的推荐效率,改进算法的离线计算方式,提前计量物品与物品之间的同好度表,同时,随机抽取百多名用户,计算新旧算法平均耗时表和数据量时间比指标表.实验表明,改进的算法具有明显的效率优势,不仅运算速度提高,运算结果可以重复使用,还提高了算法耗时的稳定性.算法拓展可用于商品的同好推荐,计算两物品之间的关联度,分析事件发生的影响因素等.
An improved recommendation algorithm was proposed based on matrix model to improve the computing speed and stability of the traditional algorithms.With the data about the information of downloaded books for nearly a year by mobile community users,the recommendation efficiency,average time consuming and the ratio between data size and time of the proposed algorithm were analyzed with the comparison of the traditional recommendation algorithms.The analysis results show that the recommendation efficiency is increased obviously,and the computing speed and the stability of time consuming are also improved; moreover,the form of offline calculation is modified in the proposed algorithm,and enthusiast table between different goods is pre-calculated in offline form.The proposed algorithm can be applied in enthusiast recommendation of product,computing the correlation between the two items and analyzing the impact of factors such as events.
出处
《武汉工程大学学报》
CAS
2014年第8期75-78,共4页
Journal of Wuhan Institute of Technology
基金
福建省B类项目(JB13197)
关键词
同好度
同好推荐算法
矩阵模型
数据挖掘
关联度
enthusiasts degree
enthusiasts recommendation algorithm
matrix model
data mining
correlation