余弦相似度加权的Slope One协同过滤算法研究

Abstract

针对slope one协同过滤算法中存在的数据稀疏性问题展开研究。提出一种基于余弦相似度加权的协同过滤算法(COSLOPE算法)。用加权slope one算法填充稀疏的评分矩阵后利用cosine算法计算用户之间的相似度，得出目标用户的近邻矩阵。通过近邻矩阵中拥有评分记录的用户来预测目标用户的项目评分，并进行推荐。该算法通过MovieLens数据集验证，MAE、RMSE 和MSE的值均优于传统Slope One算法。COSLOPE算法在有效解决数据稀疏性的同时亦提高了传统推荐算法的准确度并降低了算法响应时间。 In this paper, we propose a collaborative filtering algorithm based on cosine similarity weight (COSLOPE algorithm). The similarity between the users is calculated by the cosine algorithm; the weights are determined according to the similarity degree and the scoring matrix is filled in order to establish the nearest neighbor set with high similarity to the object user. The nearest neighbor set of the nearest neighbor set is to predict the target user’s project grade and make recommendations. The algorithm is validated by the MovieLens dataset, and the values of MAE, RMSE and MSE are superior to the traditional Slope One algorithm. COSLOPE algorithm is not only in the effective solution of data sparseness, but also improve the accuracy of the traditional recommendation algorithm and reduce the algorithm response time.