从余弦相似度值聚类

本文介绍了从余弦相似度值聚类的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！ 问题描述

我从一组URL中提取了单词并计算了每个URL内容之间的余弦相似度，并且我将值标准化为0-1(使用Min-Max).现在我需要基于余弦相似度对URL进行聚类值以找出相似的URL.哪种聚类算法最合适?.请向我建议一种动态聚类方法，因为它会很有用，因为我可以根据需要增加URL的数量，而且也会更自然.请对我进行纠正感觉我在以错误的方式取得进展.感谢您的期待.

I have extracted words from a set of URLs and calculated cosine similarity between each URL's contents.And also I have normalized the values between 0-1(using Min-Max).Now i need to cluster the URLs based on cosine similarity values to find out similar URLs.which clustering algorithm will be most suitable?.Please suggest me a Dynamic clustering method because it will be useful since i could increase number of URL's on demand and also it will be more natural.Please correct me if you feel i'm making the progress in a wrong way.Thanks in anticipation.

推荐答案

K-means聚类可用于在线学习，您只需要事先选择聚类的数量即可.另外，我认为您不应该对数据进行规范化，因为余弦已经提供了[0:1]范围内的值.您的Min-Max规范化可能会导致信息丢失.

K-means clustering can be used for online learning, you just need to select the number of clusters a priori. Also, I think you shouldn't normalize your data, because cosine already provides values in the range [0:1]. Your Min-Max normalization could lead to information loss.