数据挖掘（5）模型调优（网格搜索）

使用网格搜索法对5个模型进行调优（调参时采用五折交叉验证的方式），并进行模型评估

可以使用sklearn中的网格搜索

from sklearn.model_selection import GridSearchCV

clf = LogisticRegression(C=1.0, max_iter=1000).fit(train_data, train_label)
parameters = {'C':[1.0,2.0,3.0,4.0,5.0], 'max_iter': [100,200,500,1000,1500,2000]}
grid = GridSearchCV(clf, parameters, cv=5)
grid = grid.fit(train_data, train_label)
grid_test = grid.predict(test_data)
grid_train = grid.predict(train_data)

from sklearn import metrics
import pandas as pd
import numpy as np
from sklearn.model_selection import StratifiedKFold
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn import svm
from xgboost.sklearn import XGBClassifier
from sklearn.feature_selection import RFE
from sklearn.metrics import roc_curve, auc
import matplotlib.pyplot as plt
from sklearn.model_selection import GridSearchCV

def LR_classifier(train_data, train_label, test_data, test_label):
    clf = LogisticRegression(C=1.0, max_iter=1000).fit(train_data, train_label)
    parameters = {'C':[1.0,2.0,3.0,4.0,5.0], 'max_iter': [100,200,500,1000,1500,2000]}
    grid = GridSearchCV(clf, parameters, cv=5)
    grid_dtc = grid.fit(train_data, train_label)
    grid_test = grid_dtc.predict(test_data)
    grid_train = grid_dtc.predict(train_data)
    return grid_test, grid_train

def svm_classifier(train_data, train_label, test_data, test_label):
    clf = svm.SVC(C=1.0, kernel='linear', gamma=20).fit(train_data, train_label)
    parameters = {'C':[1.0,2.0,3.0], 'gamma':[5,10,15, 20, 25]}
    grid = GridSearchCV(clf, parameters, cv=5)
    grid_dtc = grid.fit(train_data, train_label)
    grid_test = grid_dtc.predict(test_data)
    grid_train = grid_dtc.predict(train_data)
    return grid_test, grid_train

def dt_classifier(train_data, train_label, test_data, test_label):
    clf = DecisionTreeClassifier(max_depth=5).fit(train_data, train_label)
    parameters = {'max_depth':[2,5,8,10,15]}
    grid = GridSearchCV(clf, parameters, cv=5)
    grid_dtc = grid.fit(train_data, train_label)
    grid_test = grid_dtc.predict(test_data)
    grid_train = grid_dtc.predict(train_data)
    return grid_test, grid_train

def rf_classifier(train_data, train_label, test_data, test_label):
    clf = RandomForestClassifier(n_estimators=8, random_state=5, max_depth=6, min_samples_split=2).fit(train_data, train_label)
    parameters = {'n_estimators':[3,5,8,10,14], 'random_state':[2,3,5,7,9],'max_depth':[5,6,8,9,10,15],'min_samples_split':[2,3,4,5,6]}
    grid = GridSearchCV(clf, parameters, cv=5)
    grid_dtc = grid.fit(train_data, train_label)
    grid_test = grid_dtc.predict(test_data)
    grid_train = grid_dtc.predict(train_data)
    return grid_test, grid_train

def xgb_classifier(train_data, train_label, test_data, test_label):
    clf = XGBClassifier(n_estimators=8,learning_rate= 0.25, max_depth=20,subsample=1,gamma=13, seed=1000,num_class=1).fit(train_data, train_label)
    parameters = {'n_estimators':[3,5,8,10,14], 'learning_rate':[0.1,0.2,0.25,0.3,0.35,0.4],'max_depth':[5,10,15,20,25],'gamma':[6,9,12,13,15],'seed':[500,1000,1500]}
    grid = GridSearchCV(clf, parameters, cv=5)
    grid_dtc = grid.fit(train_data, train_label)
    grid_test = grid_dtc.predict(test_data)
    grid_train = grid_dtc.predict(train_data)
    return grid_test, grid_train

模型	Accuracy	Precision	Recall	F1_score	AUC	ROC
Logistic Regression	train:0.7913 test:0.787	train:0.7351 test:0.7195	train:0.2668, test:0.2552	train:0.3915, test:0.3759	train:0.6173, test:0.6105
Support Vector Machine	train:0.7793, test:0.7762	train:0.8025, test:0.7783	train:0.1632, test:0.1554	train:0.2712, test:0.2588	train:0.5748, test:0.5702
Decision Tree	train:0.78, test:0.7732	train:0.8241, test:0.7474	train:0.1611, test:0.1537	train:0.2632, test:0.2483	train:0.5746, test:0.5676
Random Forest	train:0.8311, test:0.778	train:0.9326, test:0.7077	train:0.3504, test:0.2058	train:0.502, test:0.3172	train:0.6716, test:0.5881
XGBoost	train:0.839, test:0.7843	train:0.8456, test:0.6484	train:0.4402, test:0.3127	train:0.5786, test:0.4215	train:0.7067, test:0.6278