改进的基于用户的协同过滤算法

在推荐系统中,协同过滤算法是应用最广泛和最成熟的推荐算法之一。但是传统的协同过滤算法,在计算用户之间的相似度和评分推荐两个指标上存在着很多不足之处。通过决策树策略找寻了评分和共现值之间的规则,有效的改善了Salton相似度的准确性。同时,根据艾宾浩斯遗忘规律得到启发,引入了时间模型作为评分的权重,有效的解决了用户的兴趣迁移。在仿真实验中,测试了在不同邻居个数下传统算法和改进算法的平均绝对误差。实验证明,改进的协同过滤算法能够降低预测评分的平均绝对误差,提高推荐的准确率。

Improved Collaborative Filtering Algorithm Based on User

Collaborative filtering algorithm is one of the most widely used and most mature recommendation algorithms in recommender systems. But the traditional collaborative filtering algorithm has many shortcomings in calculating the similarity between users and the recommendation of two indicators. In this paper,the decision tree strategy is used to find the rules between the score and the total present value,which can effectively improve the accuracy of the Salton similarity. At the same time,according to Ebbinghaus's forgetting rule,the time model is introduced as the weight of the score,which can effectively solve the migration of user's interest. In simulation experiments,the average absolute error between the traditional algorithm and the improved algorithm is tested under different neighbor numbers. Experi-mental results show that the improved collaborative filtering algorithm can reduce the average absolute error of predic-tion score and improve the accuracy of recommendation.