转移概率矩阵:马尔科夫链 { X n , n ≥ 0 } \left\{X_{n}, n \geq 0\right\} {
Xn,n≥0}在时刻m处于状态i的条件下,在时刻m+n转移的状态j的条件概率记为n步转移概率,记为 { X m + n = j ∣ , X m = i } \left\{X_{m+n} = j|, X_m=i\right\} {
Xm+n=j∣,Xm=i}
显然有结论: ∑ j ∈ E P ( X m + n = j ∣ X m = i ) = 1 \sum_{j \in E} P\left(X_{m+n}=j \mid X_{m}=i\right)=1 ∑j∈EP(Xm+n=j∣Xm=i)=1
闵可夫斯基距离: d ( x i , x j ) = ( ∑ k = 1 d ∣ x i k − x j k ∣ q ) 1 q d\left(x_{i}, x_{j}\right)=\left(\sum_{k=1}^{d}\left|x_{i k}-x_{j k}\right|^{q}\right)^{\frac{1}{q}} d(xi,xj)=(∑k=1d∣xik−xjk∣q)q1
当q=1时,称绝对距离,当q=2,称欧式距离
马氏距离: d i j 2 ( M ) = ( X i − X j ) ′ Σ − 1 ( X i − X j ) d_{i j}^{2}(M)=\left(X_{i}-X_{j}\right)^{\prime} \Sigma^{-1}\left(X_{i}-X_{j}\right) dij2(M)=(Xi−Xj)′Σ−1(Xi−Xj)
R型聚类:对变量进行分类处理,距离由变量相似性来度量
用相关系数或者夹角余弦来评估
夹角余弦: cos θ i j = ∑ k = 1 p x i k x j k ∑ k = 1 p x i k 2 ∑ k = 1 p x j k 2 \cos \theta_{i j}=\frac{\sum_{k=1}^{p} x_{i k} x_{j k}}{\sqrt{\sum_{k=1}^{p} x_{i k}^{2}} \sqrt{\sum_{k=1}^{p} x_{j k}^{2}}} cosθij=∑k=1pxik2∑k=1pxjk2∑k=1pxikxjk