摘要
在网络图模型构建推荐系统中采用随机游走算法有效且易于实现,但存在时间复杂度高的问题.本文通过转移概率矩阵方法,将需要多次迭代的随机游走算法转换成一次矩阵求逆的过程,并使用LU分解方法解决实际应用中大规模数据矩阵求逆对系统资源高消耗的问题,通过对比实验证明了本文方法的有效性.
Random walk algorithm can be effective to adopt and easy to practice when constructing a recommendation system in network graph model,but it brings forth a rather high degree of time complexity.Through the approach of transition probability matrix,the present paper simplifies the random walk algorithm with multiple iterations into a process of one-step matrix inversion.Also in the practical usage,LU decomposition is valid to tackle the problem of high resource consumption caused by matrix inversion of Large Scale Data.Lastly,a comparative experiment proves the effectiveness of our method.
出处
《玉林师范学院学报》
2016年第2期114-118,123,共6页
Journal of Yulin Normal University
基金
广西高校科学技术研究项目
项目编号:KY2015LX300
KY2015YB241
2013LX112
关键词
推荐算法
网络图模型
随机游走算法
LU分解法
recommendation algorithm
network graph model
random walk algorithm
LU decomposition