机器学习sklearn（92）：算法实例（49）分类（30）XGBoost（六）XGBoost应用中的其他问题

1 过拟合：剪枝参数与回归模型调参

 

 

 

 

 

 

 

 

 

 

 

 

dfull = xgb.DMatrix(X,y)
param1 = {'silent':True #并非默认
         ,'obj':'reg:linear' #并非默认
         ,"subsample":1
         ,"max_depth":6
         ,"eta":0.3
         ,"gamma":0
         ,"lambda":1
         ,"alpha":0
         ,"colsample_bytree":1
         ,"colsample_bylevel":1
         ,"colsample_bynode":1
         ,"nfold":5}
num_round = 200
time0 = time()
cvresult1 = xgb.cv(param1, dfull, num_round)
print(datetime.datetime.fromtimestamp(time()-time0).strftime("%M:%S:%f"))
fig,ax = plt.subplots(1,figsize=(15,10))
#ax.set_ylim(top=5)
ax.grid()
ax.plot(range(1,201),cvresult1.iloc[:,0],c="red",label="train,original")
ax.plot(range(1,201),cvresult1.iloc[:,2],c="orange",label="test,original")
ax.legend(fontsize="xx-large")
plt.show()

 

 

 

param1 = {'silent':True
         ,'obj':'reg:linear'
         ,"subsample":1
         ,"max_depth":6
         ,"eta":0.3
         ,"gamma":0
         ,"lambda":1
         ,"alpha":0
         ,"colsample_bytree":1
         ,"colsample_bylevel":1
         ,"colsample_bynode":1
         ,"nfold":5}
num_round = 200
time0 = time()
cvresult1 = xgb.cv(param1, dfull, num_round)
print(datetime.datetime.fromtimestamp(time()-time0).strftime("%M:%S:%f"))
fig,ax = plt.subplots(1,figsize=(15,8))
ax.set_ylim(top=5)
ax.grid()
ax.plot(range(1,201),cvresult1.iloc[:,0],c="red",label="train,original")
ax.plot(range(1,201),cvresult1.iloc[:,2],c="orange",label="test,original")
param2 = {'silent':True
         ,'obj':'reg:linear'
         ,"nfold":5}
param3 = {'silent':True
         ,'obj':'reg:linear'
         ,"nfold":5}
time0 = time()
cvresult2 = xgb.cv(param2, dfull, num_round)
print(datetime.datetime.fromtimestamp(time()-time0).strftime("%M:%S:%f"))
time0 = time()
cvresult3 = xgb.cv(param3, dfull, num_round)
print(datetime.datetime.fromtimestamp(time()-time0).strftime("%M:%S:%f"))
ax.plot(range(1,201),cvresult2.iloc[:,0],c="green",label="train,last")
ax.plot(range(1,201),cvresult2.iloc[:,2],c="blue",label="test,last")
ax.plot(range(1,201),cvresult3.iloc[:,0],c="gray",label="train,this")
ax.plot(range(1,201),cvresult3.iloc[:,2],c="pink",label="test,this")
ax.legend(fontsize="xx-large")
plt.show()

在这里，为大家提供我调出来的结果，供大家参考：

#默认设置
param1 = {'silent':True
         ,'obj':'reg:linear'
         ,"subsample":1
         ,"max_depth":6
         ,"eta":0.3
         ,"gamma":0
         ,"lambda":1
         ,"alpha":0
         ,"colsample_bytree":1
         ,"colsample_bylevel":1
         ,"colsample_bynode":1
         ,"nfold":5}
num_round = 200
time0 = time()
cvresult1 = xgb.cv(param1, dfull, num_round)
print(datetime.datetime.fromtimestamp(time()-time0).strftime("%M:%S:%f"))
fig,ax = plt.subplots(1,figsize=(15,8))
ax.set_ylim(top=5)
ax.grid()
ax.plot(range(1,201),cvresult1.iloc[:,0],c="red",label="train,original")
ax.plot(range(1,201),cvresult1.iloc[:,2],c="orange",label="test,original") 

#调参结果1
param2 = {'silent':True
         ,'obj':'reg:linear'
         ,"subsample":1
         ,"eta":0.05
         ,"gamma":20
         ,"lambda":3.5
         ,"alpha":0.2
         ,"max_depth":4
         ,"colsample_bytree":0.4
         ,"colsample_bylevel":0.6
         ,"colsample_bynode":1
         ,"nfold":5} 

#调参结果2
param3 = {'silent':True
         ,'obj':'reg:linear'
         ,"max_depth":2
         ,"eta":0.05
         ,"gamma":0
         ,"lambda":1
         ,"alpha":0
         ,"colsample_bytree":1
         ,"colsample_bylevel":0.4
         ,"colsample_bynode":1
         ,"nfold":5}
time0 = time()
cvresult2 = xgb.cv(param2, dfull, num_round)
print(datetime.datetime.fromtimestamp(time()-time0).strftime("%M:%S:%f"))
ax.plot(range(1,201),cvresult2.iloc[:,0],c="green",label="train,final")
ax.plot(range(1,201),cvresult2.iloc[:,2],c="blue",label="test,final")
ax.legend(fontsize="xx-large")
plt.show()

 

 

2 XGBoost模型的保存和调用

2.1 使用Pickle保存和调用模型

 

 

 

import pickle
dtrain = xgb.DMatrix(Xtrain,Ytrain) #设定参数，对模型进行训练
param = {'silent':True
         ,'obj':'reg:linear'
         ,"subsample":1
         ,"eta":0.05
         ,"gamma":20
         ,"lambda":3.5
         ,"alpha":0.2
         ,"max_depth":4
         ,"colsample_bytree":0.4
         ,"colsample_bylevel":0.6
         ,"colsample_bynode":1}
num_round = 180
bst = xgb.train(param, dtrain, num_round) #保存模型
pickle.dump(bst, open("xgboostonboston.dat","wb"))
#注意，open中我们往往使用w或者r作为读取的模式，但其实w与r只能用于文本文件，当我们希望导入的不是文本文件，而
是模型本身的时候，我们使用"wb"和"rb"作为读取的模式。其中wb表示以二进制写入，rb表示以二进制读入
#看看模型被保存到了哪里？
import sys
sys.path
#重新打开jupyter lab
from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split as TTS
from sklearn.metrics import mean_squared_error as MSE
import pickle
import xgboost as xgb
data = load_boston()
X = data.data
y = data.target
Xtrain,Xtest,Ytrain,Ytest = TTS(X,y,test_size=0.3,random_state=420) #注意，如果我们保存的模型是xgboost库中建立的模型，则导入的数据类型也必须是xgboost库中的数据类型
dtest = xgb.DMatrix(Xtest,Ytest) #导入模型
loaded_model = pickle.load(open("xgboostonboston.dat", "rb"))
print("Loaded model from: xgboostonboston.dat") #做预测
ypreds = loaded_model.predict(dtest)
from sklearn.metrics import mean_squared_error as MSE, r2_score
MSE(Ytest,ypreds)
r2_score(Ytest,ypreds)

2.2 使用Joblib保存和调用模型 

 

 

 

bst = xgb.train(param, dtrain, num_round)
import joblib
#同样可以看看模型被保存到了哪里
joblib.dump(bst,"xgboost-boston.dat")
loaded_model = joblib.load("xgboost-boston.dat")
ypreds = loaded_model.predict(dtest)
MSE(Ytest, ypreds)
r2_score(Ytest,ypreds) #使用sklearn中的模型
from xgboost import XGBRegressor as XGBR
bst = XGBR(n_estimators=200
           ,eta=0.05,gamma=20
           ,reg_lambda=3.5
           ,reg_alpha=0.2
           ,max_depth=4
           ,colsample_bytree=0.4
           ,colsample_bylevel=0.6).fit(Xtrain,Ytrain)
joblib.dump(bst,"xgboost-boston.dat")
loaded_model = joblib.load("xgboost-boston.dat") #则这里可以直接导入Xtest
ypreds = loaded_model.predict(Xtest)
MSE(Ytest, ypreds)

3 分类案例：XGB中的样本不均衡问题 

1. 导库，创建样本不均衡的数据集 

import numpy as np
import xgboost as xgb
import matplotlib.pyplot as plt
from xgboost import XGBClassifier as XGBC
from sklearn.datasets import make_blobs
from sklearn.model_selection import train_test_split as TTS
from sklearn.metrics import confusion_matrix as cm, recall_score as recall, roc_auc_score
as auc
class_1 = 500 #类别1有500个样本
class_2 = 50 #类别2只有50个
centers = [[0.0, 0.0], [2.0, 2.0]] #设定两个类别的中心
clusters_std = [1.5, 0.5] #设定两个类别的方差，通常来说，样本量比较大的类别会更加松散
X, y = make_blobs(n_samples=[class_1, class_2],
                  centers=centers,
                  cluster_std=clusters_std,
                  random_state=0, shuffle=False)
Xtrain, Xtest, Ytrain, Ytest = TTS(X,y,test_size=0.3,random_state=420) (y == 1).sum() / y.shape[0]

2. 在数据集上建模：sklearn模式 

#在sklearn下建模#
clf = XGBC().fit(Xtrain,Ytrain)
ypred = clf.predict(Xtest)
clf.score(Xtest,Ytest)
cm(Ytest,ypred,labels=[1,0])
recall(Ytest,ypred)
auc(Ytest,clf.predict_proba(Xtest)[:,1])
#负/正样本比例
clf_ = XGBC(scale_pos_weight=10).fit(Xtrain,Ytrain)
ypred_ = clf_.predict(Xtest)
clf_.score(Xtest,Ytest)
cm(Ytest,ypred_,labels=[1,0])
recall(Ytest,ypred_)
auc(Ytest,clf_.predict_proba(Xtest)[:,1])
#随着样本权重逐渐增加，模型的recall,auc和准确率如何变化？
for i in [1,5,10,20,30]:
    clf_ = XGBC(scale_pos_weight=i).fit(Xtrain,Ytrain)
    ypred_ = clf_.predict(Xtest)
    print(i)
    print("\tAccuracy:{}".format(clf_.score(Xtest,Ytest)))
    print("\tRecall:{}".format(recall(Ytest,ypred_)))
    print("\tAUC:{}".format(auc(Ytest,clf_.predict_proba(Xtest)[:,1])))

3. 在数据集上建模：xgboost模式 

dtrain = xgb.DMatrix(Xtrain,Ytrain)
dtest = xgb.DMatrix(Xtest,Ytest) #看看xgboost库自带的predict接口
param= {'silent':True,'objective':'binary:logistic',"eta":0.1,"scale_pos_weight":1}
num_round = 100
bst = xgb.train(param, dtrain, num_round)
preds = bst.predict(dtest) #看看preds返回了什么？
preds
#自己设定阈值
ypred = preds.copy()
ypred[preds > 0.5] = 1
ypred[ypred != 1] = 0 #写明参数
scale_pos_weight = [1,5,10]
names = ["negative vs positive: 1"
         ,"negative vs positive: 5"
         ,"negative vs positive: 10"] #导入模型评估指标
from sklearn.metrics import accuracy_score as accuracy, recall_score as recall, 
roc_auc_score as auc
for name,i in zip(names,scale_pos_weight):
    param= {'silent':True,'objective':'binary:logistic'
           ,"eta":0.1,"scale_pos_weight":i}
    clf = xgb.train(param, dtrain, num_round)
    preds = clf.predict(dtest)
    ypred = preds.copy()
    ypred[preds > 0.5] = 1
    ypred[ypred != 1] = 0
    print(name)
    print("\tAccuracy:{}".format(accuracy(Ytest,ypred)))
    print("\tRecall:{}".format(recall(Ytest,ypred)))
    print("\tAUC:{}".format(auc(Ytest,preds)))
#当然我们也可以尝试不同的阈值
for name,i in zip(names,scale_pos_weight):
    for thres in [0.3,0.5,0.7,0.9]:
        param= {'silent':True,'objective':'binary:logistic'
               ,"eta":0.1,"scale_pos_weight":i}
        clf = xgb.train(param, dtrain, num_round)
        preds = clf.predict(dtest)
        ypred = preds.copy()
        ypred[preds > thres] = 1
        ypred[ypred != 1] = 0
        print("{},thresholds:{}".format(name,thres))
        print("\tAccuracy:{}".format(accuracy(Ytest,ypred)))
        print("\tRecall:{}".format(recall(Ytest,ypred)))
        print("\tAUC:{}".format(auc(Ytest,preds)))

 

 

4 XGBoost类中的其他参数和功能