核K-Means聚类在Folksonomy标签模糊和冗余中的应用

Application of K-Means clustering of kernel to ambiguity and redundancy of tag in Folksonomy

现有的Folksonomy标签推荐系统中,标签模糊会导致系统推荐不准确,并且影响用户建模的准确性,而标签冗余妨碍了对系统的评估。利用K-Means聚类结果抽取模糊和冗余标签时,聚类效果较差导致抽取不准确。提出使用核K-Means聚类处理标签模糊和冗余,通过非线性映射能够较好地分辨、提取并放大样本中有用的特征,提高抽取模糊标签和冗余标签的准确度。实验结果表明:核K-Means聚类对标签和资源的聚类效果更好,抽取的模糊标签和冗余标签也更准确。

Ambiguity of tag may give a false impression of success when the recommended tags ofter little utility. Redundancy of tag can hamper the effort to judge recommendations as well. When using K-Means clustering to deal with this problem, the extraction of ambiguity tags and redundancy tags was inaccurate because the clustering effect was ineffective. Therefore, the K-Means clustering of kernel algorithm was used to deal with the problem of ambiguity and redundancy on tags. This approach improved the clustering effect because it could identify, extract and enlarge useful features of the sample by non- linear mapping. The experimental results show that, the K-Means clustering of kernel algorithm has better performance in the clustering of tag and resource, and the extraction of ambiguity tag and redundancy tag is more accurate.