基于核估计的电子商务协作过滤方法

The Collaborative Filtering Method Based on Kernel Estimation for E-Commerce

论文对电子商务推荐技术进行了讨论,针对电子商务推荐技术存在的问题,提出了一种基于核估计的协作过滤方法,首先将推荐项的评分设定为其他人对推荐项评分的平均值,然后利用主成分分析进行降维,最后采用基于核估计的思想对推荐项进行预测评分。实验表明,该方法可以有效解决在用户评分数据极端稀疏情况下传统相似性度量方法存在的问题,大幅提高推荐系统的推荐质量。

This article discusses recommendation technology about E-Commerce and aims at the problem of recommendation technology about E-Commerce.it gives a collaborative filtering method based on kernel estimation.This method first gives the rating of not rated item as the average rating and use primary component analysis to do dimensionality reduction,at last uses the method of kernel estimation to predict.The results of experiment show this method could efficiently improve the extreme sparsity of user rating data, provides better recommendation results.