基于用户习惯偏好相似度的Slope One推荐算法

A Novel Recommendation Algorithm with Slope One Based on User Preference Similarity

数据稀疏是协同过滤预测精度的一个重要影响因素。Slope One算法使用简单的线性回归模型解决该问题,但它只使用评分数据做计算,未考虑相似性。提出一种基于用户习惯偏好相似度的Slope One算法(UPS Slope One)。UPS Slop eOne首先基于用户习惯偏好聚类,得到三组不同偏好的用户,然后分别计算各组评分偏差,计算时将用户习惯偏好相似度融入其中,最后使用线性回归模型预测评分。在MovieLens数据集上的实验表明,该算法可得到更高的推荐质量、预测准确性和稳定性。

Data sparsity is a main factor affecting the prediction accuracy of collaborative filtering. Slope One algorithm uses simple lin-ear regression model to solve this problem. It uses rating data to do calculation without considering the similarity. In this paper,we pro-posed an improved Slope One algorithm based on user preference similarity. First,we obtain three groups of users with different prefer-ences based on the user preference clustering,and then calculate the deviation of each group,we add the user preference similarity to deviation calculation,finally use linear regression model to predict the rating. Experiments on the MovieLens data set show that the pro-posed algorithm can achieve better recommendation quality and prediction accuracy. Meanwhile,the stability of the algorithm is also relatively satisfying.