多分类任务中每个类Acc、Specificity、Sensitivity、Precision的计算

虽然sklearn.metrics中带有许多计算分类效果的工具，但是不全，下面分享四个基本分类指标，其他的指标也可以照样编写

from sklearn.metrics import confusion_matrix#混淆矩阵
import numpy as np

def sen(Y_test,Y_pred,n):#n为分类数
    
    sen = []
    con_mat = confusion_matrix(Y_test,Y_pred)
    for i in range(n):
        tp = con_mat[i][i]
        fn = np.sum(con_mat[i,:]) - tp
        sen1 = tp / (tp + fn)
        sen.append(sen1)
        
    return sen

def pre(Y_test,Y_pred,n):
    
    pre = []
    con_mat = confusion_matrix(Y_test,Y_pred)
    for i in range(n):
        tp = con_mat[i][i]
        fp = np.sum(con_mat[:,i]) - tp
        pre1 = tp / (tp + fp)
        pre.append(pre1)
        
    return pre

def spe(Y_test,Y_pred,n):
    
    spe = []
    con_mat = confusion_matrix(Y_test,Y_pred)
    for i in range(n):
        number = np.sum(con_mat[:,:])
        tp = con_mat[i][i]
        fn = np.sum(con_mat[i,:]) - tp
        fp = np.sum(con_mat[:,i]) - tp
        tn = number - tp - fn - fp
        spe1 = tn / (tn + fp)
        spe.append(spe1)
    
    return spe

def ACC(Y_test,Y_pred,n):
    
    acc = []
    con_mat = confusion_matrix(Y_test,Y_pred)
    for i in range(n):
        number = np.sum(con_mat[:,:])
        tp = con_mat[i][i]
        fn = np.sum(con_mat[i,:]) - tp
        fp = np.sum(con_mat[:,i]) - tp
        tn = number - tp - fn - fp
        acc1 = (tp + tn) / number
        acc.append(acc1)
        
    return acc