基于用户特征分解的协同过滤冷启动解决算法

协同过滤中的冷启动问题是个性化推荐系统的一个难点,本文提出了一种基于用户特征分解的解决算法。首先构建了用户表示矩阵同时为保证不同用户评价的可比性,把每个用户的评价向量进行标归一化处理得到标准用户评价矩阵然后将两者进行合成得到用户特征对资源的标准评价矩阵。当新用户表示为用户特征向量,与特征评价矩阵合成得到新用户的预测评价向量,从而进行个性化推荐。基于MovieLens数据集进行的实验表明,该算法可以一定程度上解决系统冷启动问题,提高系统推荐质量。该算法还可以很方便地推广解决新资源的冷启动问题。

The Explicit Expressions of Level- k(r1 ,r2 ,… ,rk) Circulant Matrices and Their Some Properties

In this paper,we give the explicit expressions of level k (r 1,r 2,…,r k) circulant matrices of order n 1n 2…n k,and the explicit expressions for the eigenvalues,the determinants and the inverse matrices of the kind level k (r 1,r 2,…,r k) circulant matrices are derived,and it is also proved that the sort of matrices are diagonalizable.

The Explicit Expressions of Level-k Circulant Matrices of Type(n_1,n_2.…,n_k) and Their Some Properties

In this paper, we give the explicit expressions of level-k circulant matrices of type (n1,n2,…nk) and of order n1n2…nk,and the explicit expressions for the eigenvalues,the determinants and the inverse matrices of the kind level-k circulant matrices are derived,and it is also proved that the sort matrices are unitarily diagonalizable.