在Scipy层次聚类中对树状图进行修剪 [英] Pruning dendrogram at levels in Scipy Hierarchical Clustering

问题描述

我有很多数据点，这些数据点使用Scipy层次聚类以以下方式聚类.假设我要修剪"1500"级别的树状图?怎么做? (我尝试使用'p'参数，但这不是我期望的值)

I have lot of data points which are clustered in the following way using Scipy Hierarchical Clustering. Let's say I want to prune the dendogram at level '1500'? How to do that? (I've tried using 'p' parameter and that is not what I'm expecting)

Z = dendrogram(linkage_matrix,
           truncate_mode='lastp',
           color_threshold=1,
           labels=df.session.tolist(),
           distance_sort='ascending')


plt.title("Hierachical Clustering")
plt.show()

推荐答案

如

As specified in the scipy documentation, if a cluster node is under color_threshold, then all of its descendants will be the same color (not blue). The links connecting nodes above color_threshold will be blue.

在您的示例中，为color_threshold=1.由于所有节点都在1之上，因此所有链接均为蓝色.

In your example, color_threshold=1. Since all the nodes are above 1, all of the links are blue.

尝试

Z = dendrogram(linkage_matrix,
       color_threshold=1500,
       distance_sort='ascending')

这篇关于在Scipy层次聚类中对树状图进行修剪的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！