-- coding: utf-8 --

"""

Created on Sun Dec 23 01:10:07 2018

@author: NickyChu

"""

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn import svm

from sklearn.linear_model import LogisticRegression as LR

from sklearn.linear_model import RandomizedLogisticRegression as RLR

In[121]:

df = pd.read_csv('Train.csv',header=None)#使用header选项让列名为从0开始的数列

df_new = pd.read_csv('Test.csv',header=None)

print(df.duplicated())#检查重复项

df = df.drop_duplicates()#清洗重复项

df

In[163]:

my_matrix = df

data_x=my_matrix.iloc[:,0:-1]

data_y=my_matrix.iloc[:,-1]

In[123]:

rlr = RLR() #建立随机逻辑回归模型，筛选变量

rlr.fit(data_x, data_y) #训练模型

array1 = rlr.get_support() #获取特征筛选结果，也可以通过.scores_方法获取各个特征的分数

score1 = rlr.scores_

score1

In[161]:

i = 1

for name in array1:

if name == True:

print("第{0}个特征值为有用特征值".format(i))

i = i+1

In[162]:

筛选特征值

data_x.columns = score1

df_new.columns = score1

todrop = [0]

try:

data_x.drop(todrop, inplace=True, axis=1)

df_new.drop(todrop, inplace=True, axis=1)

except:

print("No need to drop")

data_x

df_new

In[]

分组

from sklearn.cross_validation import train_test_split

from sklearn.metrics import confusion_matrix, classification_report, roc_curve, auc

from sklearn.grid_search import GridSearchCV

X1_train, X1_test, y1_train, y1_test = train_test_split(data_x,data_y,test_size=0.33)

组内预测

lr=LR(C = 1.5, penalty = 'l1',class_weight='balanced')#逻辑回归模型

lr.fit(X1_train,y1_train)#train组内模拟

print('逻辑回归模型训练结束')

print('模型的平均正确率:%s' % lr.score(X1_train,y1_train))#输出模型平均正确率（不准确）

print(lr.get_params)#输出相关信息

y1_pre_val = lr.predict(X1_test)#train组内的预测函数

输出报告

print("Classification report (training):\n {0}".format(classification_report(y1_test,y1_pre_val,target_names=["0","1"])))

让他去预测一下test组的

y_val_old = lr.predict(df_new)

print(lr.get_params)#输出相关信息

In[]

parameter tuning

运用穷尽网格搜索

from sklearn.grid_search import GridSearchCV

lr_clf_tuned = LR()

tuned_parameters=[{'penalty':['l1','l2'],

'C':[0.01,0.05,0.1,0.5,1,1.1,1.2,1.3,1.4,1.5,10],

'solver':['liblinear'],

'multi_class':['ovr']},

{'penalty':['l2'],

'C':[0.01,0.05,0.1,0.5,1,1.1,1.2,1.3,1.4,1.5,10],

'solver':['lbfgs'],

'multi_class':['ovr','multinomial']}]

lr_clf_params = {"penalty": ["l1", "l2"], "C": [1, 1.1,1.2,1.3,1.4,1.5, 1.7, 2] }

lr_clf_cv = GridSearchCV(lr_clf_tuned, tuned_parameters, cv=5)

lr_clf_cv.fit(X1_train,y1_train)

print(lr_clf_cv.best_params_)

In[]

from sklearn.ensemble import RandomForestClassifier

lr=LR( C = 10, multi_class='ovr', penalty = 'l1', solver='lbfgs')#逻辑回归模型

lr.fit(X1_train,y1_train)

print('逻辑回归模型训练结束')

print('模型的平均正确率:%s' % lr.score(X1_train,y1_train))

print(lr.get_params)

y1_pre_val = lr.predict(X1_test)#train组内的预测函数

输出报告

print("Classification report (training):\n {0}".format(classification_report(y1_test,y1_pre_val,target_names=["0","1"])))

让他去预测一下test组的

y_val_new = lr.predict(df_new)

print(lr.get_params)#输出相关信息

In[]

寻找最优Roc值

from sklearn.metrics import confusion_matrix, classification_report, roc_curve, auc

from sklearn.metrics import roc_auc_score

def FindtheC():

score = []

aucarr = []

Conclu = []

for i in range(1,100):

i = 64

play = i/1000

for j in range(1,100):

j = 67时 达到 0.699的AUC

Canshu = j

lr=LR(C = Canshu, multi_class='ovr', penalty = 'l2', solver='liblinear',class_weight={0: play, 1: 1-play})#逻辑回归模型

lr.fit(X1_train,y1_train)

print('逻辑回归模型训练结束')

print('模型的平均正确率:%s' % lr.score(X1_train,y1_train))

print(lr.get_params)

y1_pre_test1 = lr.predict(X1_test)#预测函数

auc1 = "ROC是{0}".format(roc_auc_score(y1_test, y1_pre_test1))

print(auc1)

aucarr.append(roc_auc_score(y1_test, y1_pre_test1))

Conclu.append([auc1,Canshu])

print(max(aucarr))

In[]

pre_probs = lr.predict_log_proba(data_x)

print(pre_probs[:,1])

print(pre_probs[:,0])

In[]

from sklearn.ensemble import RandomForestRegressor

from sklearn.linear_model import LogisticRegression

参数调整

x_train = X1_train

y_train = y1_train

x_test = X1_test

y_test = y1_test

def show_accuracy(a, b, tip):

acc = a.ravel() == b.ravel()

acc_rate = 100 * float(acc.sum()) / a.size

return acc_rate

Maincode

lr = LogisticRegression(penalty='l2')

lr.fit(x_train, y_train)

y_hat = lr.predict(x_test)

lr_rate = show_accuracy(y_hat, y_test, 'Logistic回归 ')

print(roc_auc_score(y_test, y_hat))

随机森林 n_estimators：决策树的个数,越多越好,不过值越大，性能就会越差,至少100

rfc = RandomForestClassifier(n_estimators=100)

rfc.fit(x_train, y_train)

y_hat = rfc.predict(x_test)

rfc_rate = show_accuracy(y_hat, y_test, '随机森林 ')

print(roc_auc_score(y_test, y_hat))

In[]

XGBoost

import xgboost as xgb

data_train = xgb.DMatrix(x_train, label=y_train)

data_test = xgb.DMatrix(x_test, label=y_test)

watch_list = [(data_test, 'eval'), (data_train, 'train')]

param = {'max_depth': 6, 'eta': 0.8, 'silent': 1, 'objective': 'binary:logistic'}

bst = xgb.train(param, data_train, num_boost_round=100, evals=watch_list)

y_hat = bst.predict(data_test)

y_hat[y_hat > 0.5] = 1

y_hat[~(y_hat > 0.5)] = 0

xgb_rate = show_accuracy(y_hat, y_test, 'XGBoost ')

print(roc_auc_score(y_test, y_hat))

In[]

ROC图所需参数

y1 = lr.predict_proba(data_x)

y1_valid_score_lr1 = lr.predict_proba(data_x)

fpr_lr1, tpr_lr1, thresholds_lr1 = roc_curve(data_y, y1_valid_score_lr1[:, 1])

roc_auc_lr1 = auc(fpr_lr1, tpr_lr1)

作ROC图

plt.plot(fpr_lr1, tpr_lr1, lw=2, alpha=.6)

plt.plot([0, 1], [0, 1], lw=2, linestyle="--")

plt.xlim([0, 1])

plt.ylim([0, 1.05])

plt.xlabel("FPR")

plt.ylabel("TPR")

plt.title("ROC curve")