论文部分内容阅读
如何挖掘用户感兴趣的内容并向其推荐相应的产品是许多商务网站关注的重点。本文提出一个用户页面兴趣度公式，据此得到一个稀疏的页面兴趣度矩阵，考虑到矩阵的稀疏性会影响预测的准确性，我们首先用稀疏矩阵中每列非零元素的平均值对该列中值为零的元素进行修正，然后用奇异值分解法对原稀疏矩阵中值为零的元素进行再次修正，进而根据修正后的矩阵用SlopeOne算法对用户的页面兴趣度进行预测并用某网站的数据验证。结果表明：与稀疏矩阵相比，用修正后的矩阵预测的准确性有较大提高。“,”How to mine visitors＇ interest information and recommend corresponding product to them is a goal that many websites pursuit. The paper proposes a users＇ page interest degree formula, then gets a sparse matrix about page interest degree. Considering sparse matrix can affect predicting accuracy, this paper first uses every column＇s not zero value＇s average to replace those zeroes, second using singular value decomposition to deal with those zero values in original sparse generator matrix and get a updated matrix, third applying Slope One Algorithm to predict visitors＇ page interest degree. At last, this paper tests the result by using a website data, experiment result is： comparing with sparse generator matrix, the updated matrix after SVD under Slope One is better in predicting user＇s page interest degree＇s accuracy.