基于邻近项目的Slope One协同过滤算法

协同过滤是个性化推荐系统中的常用技术，数据稀疏性是影响协同过滤算法预测准确度的主要因素。提出了改进的Slope One算法，在该算法中，首先根据用户历史评分计算项目间相似性，然后依据项目相似性选取当前活跃用户评价过的k个相似项目记为邻近项目集合，并计算目标项目与其邻近项目的评分偏差，最后以项目间相似性为权重，计算当前活跃用户对目标项目的评分预测值。该算法使用邻近项目进行计算，降低数据的稀疏性，同时减少了计算量。使用标准MovieLens数据集对该算法的预测结果进行验证，结果表明：相对于原算法，该算法提高了预测的准确性，与其他协同过滤算法相比，推荐准确度也有明显的提高。

Slope One collaborative filtering recommendation algorithm based on neighbor

Collaborative filtering is the commonly used technology in recommendation systems. Sparsity in the data set is the main limitation in collaborative filtering algorithm. In this paper, a new recommendation algorithm is proposed. First, the similarity between items is calculated. Then the k-nearest items are chosen to calculate the deviation between the target item and its k-nearest items. Finally, the similarity between items is employed as the weight to calculate predictive value of the target item. In the proposed algorithm, only the k-nearest items are chosen to calculate predict value, which not only ensures the accuracy of prediction but also reduces the calculated amount. Experiments on the MovieLens dataset show that the proposed algorithm gives better recommendations. It also outperforms other collaborative filtering algorithms on prediction accuracy.