【随机森林】

#导入必要的包
from sklearn.ensemble import RandomForestClassifier 
from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split,GridSearchCV,cross_val_score
from sklearn.metrics import accuracy_score
import matplotlib.pyplot as plt
import numpy as np

#导入数据集
data = load_digits()
x = data.data
y = data.target

RF = RandomForestClassifier(random_state = 66)
score = cross_val_score(RF,x,y,cv=10).mean()
print('交叉验证得分: %.4f'%score)

RF = RandomForestClassifier(random_state = 66)
score = cross_val_score(RF,x,y,cv=10).mean()
print('基尼系数得分: %.4f'%score)
RF = RandomForestClassifier(criterion = 'entropy',random_state = 66)
score = cross_val_score(RF,x,y,cv=10).mean()
print('熵得分: %.4f'%score)

###调n_estimators参数
ScoreAll = []
for i in range(10,200,10):
    DT = RandomForestClassifier(n_estimators = i,random_state = 66) #,criterion = 'entropy'
    score = cross_val_score(DT,data.data,data.target,cv=10).mean()
    ScoreAll.append([i,score])
ScoreAll = np.array(ScoreAll)

max_score = np.where(ScoreAll==np.max(ScoreAll[:,1]))[0][0] ##这句话看似很长的，其实就是找出最高得分对应的索引
print("最优参数以及最高得分:",ScoreAll[max_score])  
plt.figure(figsize=[20,5])
plt.plot(ScoreAll[:,0],ScoreAll[:,1])
plt.show()

###进一步缩小范围，调n_estimators参数
ScoreAll = []
for i in range(100,130):
    DT = RandomForestClassifier(n_estimators = i,random_state = 66)   #criterion = 'entropy',
    score = cross_val_score(DT,data.data,data.target,cv=10).mean()
    ScoreAll.append([i,score])
ScoreAll = np.array(ScoreAll)

max_score = np.where(ScoreAll==np.max(ScoreAll[:,1]))[0][0] ##这句话看似很长的，其实就是找出最高得分对应的索引
print("最优参数以及最高得分:",ScoreAll[max_score])  
plt.figure(figsize=[20,5])
plt.plot(ScoreAll[:,0],ScoreAll[:,1])
plt.show()

###粗调max_depth参数
ScoreAll = []
for i in range(10,30,3):
    DT = RandomForestClassifier(n_estimators = 117,random_state = 66,max_depth =i ) #,criterion = 'entropy'
    score = cross_val_score(DT,data.data,data.target,cv=10).mean()
    ScoreAll.append([i,score])
ScoreAll = np.array(ScoreAll)

max_score = np.where(ScoreAll==np.max(ScoreAll[:,1]))[0][0] ##这句话看似很长的，其实就是找出最高得分对应的索引
print("最优参数以及最高得分:",ScoreAll[max_score])  
plt.figure(figsize=[20,5])
plt.plot(ScoreAll[:,0],ScoreAll[:,1])
plt.show()

###调min_samples_split参数
ScoreAll = []
for i in range(2,10):
    RF = RandomForestClassifier(n_estimators = 117,random_state = 66,max_depth =16,min_samples_split = i ) #,criterion = 'entropy'
    score = cross_val_score(RF,data.data,data.target,cv=10).mean()
    ScoreAll.append([i,score])
ScoreAll = np.array(ScoreAll)

max_score = np.where(ScoreAll==np.max(ScoreAll[:,1]))[0][0] ##这句话看似很长的，其实就是找出最高得分对应的索引
print("最优参数以及最高得分:",ScoreAll[max_score])  
plt.figure(figsize=[20,5])
plt.plot(ScoreAll[:,0],ScoreAll[:,1])
plt.show()

###调min_samples_leaf参数
ScoreAll = []
for i in range(1,15,2):
    DT = RandomForestClassifier(n_estimators = 117,random_state = 66,max_depth =16,min_samples_leaf = i,min_samples_split = 2 ) 
    score = cross_val_score(DT,data.data,data.target,cv=10).mean()
    ScoreAll.append([i,score])
ScoreAll = np.array(ScoreAll)

max_score = np.where(ScoreAll==np.max(ScoreAll[:,1]))[0][0] ##这句话看似很长的，其实就是找出最高得分对应的索引
print("最优参数以及最高得分:",ScoreAll[max_score])  
plt.figure(figsize=[20,5])
plt.plot(ScoreAll[:,0],ScoreAll[:,1])
plt.show()

有条件的话，最后再进行进一步的网格搜索调参，取最优的几个参数

import time
start = time.time()

param_grid = {
  'n_estimators':np.arange(140, 150),
  'max_depth':np.arange(15, 18),
  'min_samples_leaf':np.arange(1, 8),
  'min_samples_split':np.arange(2, 5),
   'max_features':np.arange(0.1, 1)
}

rfc = RandomForestClassifier(random_state=66)
GS = GridSearchCV(rfc,param_grid,cv=10)
GS.fit(data.data,data.target)
end = time.time()
print("循环运行时间:%.2f秒"%(end-start))
print(GS.best_params_)
print(GS.best_score_)

另一种调参方式，用网格搜索来依次调参，下面是示例

#调max_features参数
param_grid = {
    'max_features':np.arange(0.1, 1)}

rfc = RandomForestClassifier(random_state=66,n_estimators = 117,max_depth = 16,min_samples_leaf =1 ,min_samples_split =4 )
GS = GridSearchCV(rfc,param_grid,cv=10)
GS.fit(data.data,data.target)
print(GS.best_params_)
print(GS.best_score_)

机器学习建模通用模板
代码参考博客

https://blog.csdn.net/u013044310/article/details/104045284?spm=1001.2101.3001.6650.1&utm_medium=distribute.pc_relevant.none-task-blog-2_defaultCTRLIST_{default-1-104045284-blog-80337103.pc_relevant_aa_2&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2}default_CTRLISTdefault-1-104045284-blog-80337103.pc_relevant_aa_2&utm_relevant_index=2