[srm] 分群分析

1 min read
Table of contents
K-means
組內變異
$$\begin{aligned} \operatorname{W}(C_k)&:=\frac{1}{|C_k|}\sum_{i,i'\in{C_k}}\operatorname{D_{Euclidean}^2}(x_i,x_{i'}) \\ &=\frac{1}{n_{c_k}}\sum_{i,i'\in{C_k}}\sum_{j=1}^p(x_{i,j}-x_{i',j})^2 \\ &=2\times\sum_{i}^N\sum_{j=1}^p(x_{i,j}-\bar{x}_{k,j})^2 \end{aligned}$$
Hierarchical Clustering
依照下列方式計算Linkage,並逐個依據當前最小的Linkage合併兩Cluster
Linkage | Inter-cluster Dissimilarity |
Complete | 比較兩群間,所有資料比的組合,取MAX的距離 |
Single | 比較兩群間,所有資料比的組合,取MIN的距離 |
Average | 比較兩群間,所有資料比的組合,取距離的MEAN |
Centroid | 計算兩群的中心點以後,取兩中心點的差 |
0
Subscribe to my newsletter
Read articles from etori directly inside your inbox. Subscribe to the newsletter, and don't miss out.
Written by

etori
etori
test bio