数据挖掘初次接触！学习代码

import pandas as pd
import numpy as np
from time import time

data=pd.read_csv("dankuan.csv",sep=';',index_col=0) #打开表
data.head() #查看前几行的数据,默认前5行
data.describe() #数据的快速统计汇总
data.g4_term_type.value_counts()
data2=data.fillna(value={'star_class':1300,'main_offer_level':0,'mon_flow_last2':0,'mon_flow_last1':0,'thrmon_flow_trend':0}).fillna(6) #缺值处理
data2.g4_term_type.value_counts(dropna=False)#计数
data2.isnull().any()#查看字段是否有空值

#pd.get_dummies

from sklearn import preprocessing

enc = preprocessing.OneHotEncoder()
a1 = data2[['g4_term_type','thrmon_flow_trend']]
enc.fit(a1)

data2_onehot=pd.DataFrame(enc.transform(a1).toarray().astype(np.bool))
# train_onehot.columns  = ['edu_class0','edu_class1','edu_class2','edu_class3','curPlan1','curPlan2','curPlan3','curPlan4']
# train2=pd.concat([train,train_onehot.set_index(train.index)],axis=1)
# test_onehot=pd.DataFrame(enc.transform(test[['edu_class', 'curPlan']]).toarray().astype(np.bool))
# test_onehot.columns  = ['edu_class0','edu_class1','edu_class2','edu_class3','curPlan1','curPlan2','curPlan3','curPlan4']
# test2=pd.concat([test,test_onehot.set_index(test.index)],axis=1)

print(enc.n_values_,enc.feature_indices_,data2_onehot.shape)

data2_onehot.columns=['data2_onehot1','data2_onehot2','data2_onehot3','data2_onehot4','data2_onehot5','data2_onehot6','data2_onehot7','data2_onehot8']#区总类别数
data3=pd.concat([data2,data2_onehot.set_index(data2.index)],axis=1)

data3.head()

names=data3.columns
names=names.drop(['is_zf_flag'])#去除无用值
data3[names].head()
names

X=data3[names]
y=data3.is_zf_flag

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)#拆分训练数据与测试数据
from sklearn.ensemble import RandomForestClassifier #算法1：自动调优算法
from sklearn.model_selection import train_test_split,GridSearchCV
rf = RandomForestClassifier(n_estimators=200,max_features=0.7,)
model = GridSearchCV(rf, param_grid={'max_depth':[1,5,10], 'min_samples_leaf':[2,5,10,20,50,100]}, cv=3)
model.fit(X_train,y_train)
y_pred = model.predict(X_test)

from sklearn.ensemble import RandomForestClassifier#算法2：随机深林算法

clf = RandomForestClassifier(n_jobs = 1)#分类型决策树；
#n_jobs：用几个处理器
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)#判定结果
y_pred.shape

print(list(zip(names,np.round(clf.feature_importances_*100,2))))#各feature的重要性
print(names[clf.feature_importances_.argmax()],np.round(clf.feature_importances_*100,2).max())

from sklearn.metrics import accuracy_score,recall_score,precision_score,f1_score
#分类准确率，提取出的正确信息条数 /样本中的信息条数，正确被检索/实际被检索,正确被检索/应该检索到

print(accuracy_score(y_test,y_pred),recall_score(y_test,y_pred))
print(precision_score(y_test,y_pred),f1_score(y_test,y_pred))

from sklearn.model_selection import cross_val_score
re_yc = cross_val_score(model, X_train, y_train, cv=10) #交叉验证用于防止模型过于复杂而引起的过拟合

from sklearn.model_selection import cross_val_score
re = cross_val_score(clf, X_train, y_train, cv=15) #交叉验证用于防止模型过于复杂而引起的过拟合

re_yc.max()

re_yc.min()

re_yc.max()-re_yc.min()

data_yc = pd.read_csv("dankuan_yc.csv",sep=';',index_col=0) #打开表
data_yc2=data_yc.fillna(value={'star_class':1300,'main_offer_level':0,'mon_flow_last2':0,'mon_flow_last1':0,'thrmon_flow_trend':0}).fillna(6) #缺值处理
a2 = data_yc2[['g4_term_type','thrmon_flow_trend']]
enc.fit(a2)
data_yc_onehot=pd.DataFrame(enc.transform(a2).toarray().astype(np.bool))
data2_onehot.columns=['data2_onehot1','data2_onehot2','data2_onehot3','data2_onehot4','data2_onehot5','data2_onehot6','data2_onehot7','data2_onehot8']#区总类别数
data3_yc=pd.concat([data_yc2,data_yc_onehot.set_index(data_yc2.index)],axis=1)
names_yc=data3_yc.columns
xxx=data3_yc[names_yc]

ans=model.predict(xxx)
pd.DataFrame(ans).set_index(xxx.index).to_csv("dkjg.csv",header=False) 

ans=clf.predict(xxx)
pd.DataFrame(ans).set_index(xxx.index).to_csv("dkjg.csv",header=False)