改进的协同过滤算法

针对传统的CF（Collaborative Filtering）算法和基于项目评分的CF算法中存在的数据稀疏、扩展性及计算效率低的问题，通过引用评价系数，对其相似性计算和推荐集的选取方法进行了改进，提出了一种改进的基于相关相似性的CF算法，产生更为准确的用户兴趣度预测，从而提高系统推荐的质量与推荐效率。对改进算法进行实验和性能对比与评价的结果表明，改进算法与传统算法相比，能显著提高推荐精度，平均绝对误差（MAE：Mean Absolute Error）为0．53～0．77。

Improvement on Correlation Similar Collaborative Filtering Algorithm

In view of the limitation of the traditional CF （ Collaborative Filtering） algorithm and the CF algorithm based on the project grading which the data are sparse, extended and the counting efficiency is low, a similar computation and the recommendation selection method based on the correlation similar CF algorithm is proposed. This method improve similar computation and the recommendation collection selection method by introducing an evaluation factor. It has a more accurate user interest forecast, and enhances the system recommendation the quality and the recommendation efficiency. The experiment and the performance analysis for the improved algorithm are made, proved that the improved algorithm can obviously increase the recommendation precision, MAE （Mean Absolute Error） is 0. 53 -0. 77.