摘要
针对传统电子商务个性化推荐系统存在的数据稀疏问题,提出一种蚁群聚类算法的个性推荐模型。利用蚁群算法的原理,找到与目标用户相似的邻居类簇,然后利用这些类簇内的用户作为基础,对目标项目中未评分项目进行预测评分,从而达到提高协同算法邻近查询的速度、降低数据稀疏性的目的;结合协同过滤思想,设计了基于时间评分的协同过滤算法,最后对上述算法进行了验证。结果表明,当最近邻居数为25的时候,目标用户的预测评分值与真实评分值的MAE差距最小,此时精度最高,说明该方法在解决数据稀疏性方面具有一定的价值。
Aiming at the problem of data sparse in the traditional e-commerce personalized recommendation system, it proposes an individual recommendation model based on ant colony clustering algorithm.Based on the principle of ant colony algorithm, it shows the process such as finding the neighbor clusters similar to the target user, and using these clusters within the user as the basis, predicting the score of the item for not score with the target project, and matrix filling.This process improves the collaborative algorithm of adjacent query speed, reduces the data sparse.Combining collaborative filtering idea with the introduction and design of the collaborative filtering algorithm based on time effect function, it verifies the algorithm improves recommendation accuracy.
出处
《机械设计与制造工程》
2017年第8期88-91,共4页
Machine Design and Manufacturing Engineering
关键词
蚁群算法
聚类技术
时间效应函数
协同过滤
预测评分
ant colony algorithm
clustering technology
time effect function
collaborative filtering
prediction score
