# -*- coding:utf-8 -*-

import pandas as pd
import pymysql
import sklearn
from sqlalchemy import create_engine
## 加上字符集参数，防止中文乱码
dbconn = pymysql.connect(
        host="127.0.0.1",
        database="yes",
        user="root",
        password="111111",
        port=3306,
        charset='utf8')
conn = create_engine('mysql+mysqldb://root:111111@localhost:3306/yes?charset=utf8')
#上面这一大段等同于conn = create_engine('mysql+mysqldb://root:111111@localhost:3306/test?charset=utf8')

# sql语句
sqlcmd = "select * from buildm"
sqlcmd1 = "select * from tobepre_N0509"

# 利用pandas 模块导入mysql数据
data = pd.read_sql(sqlcmd, dbconn)
wait = pd.read_sql(sqlcmd1, dbconn)#等待预测的数据



#shujubufen
X = data.drop([u'是否涨幅5个点',u'代码'],axis=1)
y = data[u'是否涨幅5个点']

wait1= wait.drop([u'代码'],axis=1)

#标准化  归一化#
from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler(feature_range=(0,1))
scaler.fit(X)
scaler.fit(wait1)
print X.shape

X.fillna(value=0)
wait1.fillna(value=0)

#解决样本不均衡问题,大比小为10:3
from imblearn.over_sampling import RandomOverSampler
ratio = {1:6000,0:14601}
sm = RandomOverSampler(ratio=ratio,random_state=None)

# sm = SMOTE(random_state=42,m_neighbors=5,ratio=0.3)
X_res,y_res = sm.fit_sample(X,y)


#分成训练集和测试集
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X_res, y_res, test_size=0.3,random_state=None)#分测试集和训练集
#

#神经网络----------------z暂时没法用
# from sklearn.neural_network import MLPClassifier
# MLP = MLPClassifier(hidden_layer_sizes=(450,450,300),activation='tanh',solver='adam',learning_rate_init=.1,epsilon=.000000001)
# MLP.fit(X_train,y_train)
# y_pred = pd.DataFrame(MLP.predict(X_train))
# y_true = y_train
# y_pred1 = pd.DataFrame(MLP.predict(X_test))
# y_true1 = y_test
# print 'MLP'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)
#DT可用
from sklearn.tree import DecisionTreeClassifier

DT = DecisionTreeClassifier(criterion='entropy',splitter='best',min_samples_split=20,min_samples_leaf=10,random_state=13)
DT.fit(X_train,y_train)

y_pred = pd.DataFrame(DT.predict(X_train))
y_true = y_train
y_pred1 = pd.DataFrame(DT.predict(X_test))
y_true1 = y_test
# print 'DT'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)

#逻辑回归---------------------暂不可用
# from sklearn.linear_model import LogisticRegression
#
# LR = LogisticRegression(penalty='l2', dual=False, tol=0.0001, C=2, fit_intercept=True, solver='lbfgs',  multi_class='ovr',max_iter=800)
# LR.fit(X_train,y_train)
#
# y_pred = pd.DataFrame(LR.predict(X_train))
# y_true = y_train
# y_pred1 = pd.DataFrame(LR.predict(X_test))
# y_true1 = y_test
# print 'LR'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)


#支持向量机
from sklearn.svm import SVC
SVC = SVC(C=1.0,kernel='rbf',gamma='auto',tol=0.001,probability=True)
SVC.fit(X_train,y_train)

y_pred = pd.DataFrame(SVC.predict(X_train))
y_true = y_train
y_pred1 = pd.DataFrame(SVC.predict(X_test))
y_true1 = y_test
# print 'SVC'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)

#ADABOOST决策树
from sklearn.ensemble import AdaBoostClassifier
ADA = AdaBoostClassifier(base_estimator=None,n_estimators=150, learning_rate=1.0, algorithm='SAMME')
ADA.fit(X_train,y_train)
#
# y_pred = pd.DataFrame(ADA.predict(X_train))
# y_true = y_train
# y_pred1 = pd.DataFrame(ADA.predict(X_test))
# y_true1 = y_test
# print 'ADA'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)


#随机森林
from sklearn.ensemble import RandomForestClassifier
rf=RandomForestClassifier(n_estimators=150,max_depth=5)
rf.fit(X_train,y_train)
#
# y_pred = pd.DataFrame(rf.predict(X_train))
# y_true = y_train
# y_pred1 = pd.DataFrame(rf.predict(X_test))
# y_true1 = y_test
# print 'rf'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)


#梯度树
from sklearn.ensemble import GradientBoostingClassifier
gdbt = GradientBoostingClassifier(loss='exponential',n_estimators=100,max_depth=5,min_samples_leaf=8)
gdbt.fit(X_train,y_train)
#
# y_pred = pd.DataFrame(gdbt.predict(X_train))
# y_true = y_train
# y_pred1 = pd.DataFrame(gdbt.predict(X_test))
# y_true1 = y_test
# print 'gdbt'
# print sklearn.metrics.confusion_matrix(y_true, y_pred)
# print sklearn.metrics.confusion_matrix(y_true1, y_pred1)

#knn
from sklearn import neighbors
from sklearn.neighbors import KNeighborsClassifier
KNN = KNeighborsClassifier(weights='uniform',algorithm='kd_tree',n_jobs=4,p=2,n_neighbors=7)
KNN.fit(X_train,y_train)


y_pred = pd.DataFrame(KNN.predict(X_train))
y_true = y_train
y_pred1 = pd.DataFrame(KNN.predict(X_test))
y_true1 = y_test
print 'KNN'
print sklearn.metrics.confusion_matrix(y_true, y_pred)
print sklearn.metrics.confusion_matrix(y_true1, y_pred1)

#开始拼接成果了
daima = pd.DataFrame(wait[u'代码'])#前面的U千万不可少



#开始拼接成果了
y1 = pd.DataFrame(DT.predict(wait1))
y1.columns=['y_DT']
# y2 = pd.DataFrame(MLP.predict(wait1))
# y2.columns=['y_MLP']
y_SVC = pd.DataFrame(SVC.predict(wait1))
y_SVC.columns=['y_SVC']
y3 = pd.DataFrame(ADA.predict(wait1))
y3.columns=['y_ADA']
y4 = pd.DataFrame(rf.predict(wait1))
y4.columns=['y_rf']
y5 = pd.DataFrame(gdbt.predict(wait1))
y5.columns=['y_gdbt']
# y6 = pd.DataFrame(LR.predict(wait1))
# y6.columns=['y_LR']
y7 = pd.DataFrame(KNN.predict(wait1))
y7.columns=['y_KNN']
yp = pd.DataFrame(gdbt.predict_proba(wait1))
yp.columns=['yp0','yp1']

#合并所需字段
jieguo = pd.DataFrame(pd.concat([daima,y_SVC,y1,y3,y4,y5,yp['yp1']],axis=1))#这一步是合并

jieguo['depiaoshu']=jieguo['y_ADA']*0.5+jieguo['y_rf']+jieguo['y_gdbt']+jieguo['y_DT']*0.8+jieguo['y_SVC']



pd.io.sql.to_sql(jieguo,"presult0509_py",con=conn,if_exists='replace')#只能说使用createegien的方式