用sklearn训练样本数据<4>

用sklearn分析基金数据<1>
python爬虫获取基金数据<2>
数据预处理：数据清洗、生成样本数据<3>
用sklearn训练样本数据<4>
用模型进行预测及改进<5>

拿到样本数据后就可以用sklearn中的各个分类算法来做训练了，由于样本是很简单的二分类问题，特征变量也很少，像多元分类、降维这类问题都不用考虑了，我也想过拿回归算法来做看看对下月增长率的预测情况，

LinearRegression，Ridge，Lasso的MSE得分依次降低，拟合效果略有增强，但实际测试结果误差太多，应该是样本不呈线性，也没必要转成多项式回归研究了，没有参考价值也就去掉了。
import pandas as pd
import numpy as np
Allrecords=pd.read_csv('2017-12_data.csv')
Allrecords.index=Allrecords.iloc[:,0]
Allrecords['calsstype'] = Allrecords['calsstype'].replace([0],[-1])
data= Allrecords
data_train = data.iloc[:int(0.8*len(data)), :] #选取前80%为训练数据
data_test = data.iloc[int(0.8*len(data)):, :] #选取前20%为测试数据
Xtrain = data_train.iloc[:,1:6]
Ytrain_lr = data_train.iloc[:,9]
Ytest_lr = data_test.iloc[:,9]
Xtest = data_test.iloc[:,1:6]

#LinearRegression
from sklearn.linear_model import  LinearRegression
linreg = LinearRegression()
linreg.fit(Xtrain, Ytrain_lr)
from sklearn import metrics
# 用scikit-learn计算MSE
print("MSE:",metrics.mean_squared_error(Ytest_lr, y_pred_lr))
# 用scikit-learn计算RMSE
print("RMSE:",np.sqrt(metrics.mean_squared_error(Ytest_lr, y_pred_lr)))
y_pred_lr=pd.DataFrame(y_pred_lr,index=newindex)
#Ridge
from sklearn.linear_model import Ridge
ridge = Ridge(alpha=1)
ridge.fit(Xtrain, Ytrain_lr)
print(linreg.intercept_)
print(linreg.coef_)
#用scikit-learn选择Ridge回归超参数α
from sklearn.linear_model import RidgeCV
ridgecv = RidgeCV(alphas=[0.01, 0.1, 0.5, 1, 3, 5, 7, 10, 20, 100])
ridgecv.fit(Xtrain, Ytrain_lr)
print(ridgecv.alpha_)
y_pred_ridge = ridgecv.predict(Xtest)
# 用scikit-learn计算MSE
print("MSE:",metrics.mean_squared_error(Ytest_lr, y_pred_ridge))
# 用scikit-learn计算RMSE
print("RMSE:",np.sqrt(metrics.mean_squared_error(Ytest_lr, y_pred_ridge)))
#Lasso
from sklearn.linear_model import Lasso
ridge = Lasso(alpha=1)
ridge.fit(Xtrain, Ytrain_lr)
print(linreg.intercept_)
print(linreg.coef_)
#用scikit-learn选择Ridge回归超参数α
from sklearn.linear_model import LassoCV
ridgecv = LassoCV(alphas=[0.01, 0.1, 0.5, 1, 3, 5, 7, 10, 20, 100])
ridgecv.fit(Xtrain, Ytrain_lr)
print(ridgecv.alpha_)
y_pred_lasso = ridgecv.predict(Xtest)
# 用scikit-learn计算MSE
print("MSE:",metrics.mean_squared_error(Ytest_lr, y_pred_lasso))
# 用scikit-learn计算RMSE
print("RMSE:",np.sqrt(metrics.mean_squared_error(Ytest_lr, y_pred_lasso)))

然后就是各种分类算法了，先提一下模型评估，通过分析生成样本Y值可以发现样本类型分布不均，正例数据所占比重较小，约10%，此时单看准确率已经不够了，比如全部预测为负类都有90%准确率，这时召回率(Recall)   R=TP/(TP+FN)更值得参考。

分类算法有很多，从简单开始我都试了一下

LogisticRegression，DecisionTreeClassifier，KNeighborsClassifier(KNN),GaussianNB(bayes)，以上算法结果准确率还行，但召回率很低，基本全部预测为负类也就可以排除不用了。再住复杂些分类算法走，情况就好些了，准确率都有90%+，召回率也有50%+，算法有四个，SVN和集成方法：RandomForestClassifier，AdaBoostClassifier，GradientBoostingClassifier。

具体代码如下：

import pandas as pd
Allrecords=pd.read_csv('2017-12_data.csv')
Allrecords.index=Allrecords.iloc[:,0]
Allrecords['calsstype'] = Allrecords['calsstype'].replace([0],[-1])
data= Allrecords
data_train = data.iloc[:int(0.8*len(data)), :] #选取前80%为训练数据
data_test = data.iloc[int(0.8*len(data)):, :] #选取前20%为测试数据

Ytrain = data_train.iloc[:,8]
Xtrain = data_train.iloc[:,1:6]

Ytest = data_test.iloc[:,8]
Xtest = data_test.iloc[:,1:6]

from sklearn.ensemble import RandomForestClassifier
from sklearn.grid_search import GridSearchCV
from sklearn.tree import DecisionTreeClassifier

rf1 = RandomForestClassifier(n_estimators= 50,max_depth=14,
                                  oob_score=True, random_state=10)
rf1.fit(Xtrain,Ytrain)
print(rf1.oob_score_)
y_predprob = rf1.predict_proba(Xtrain)[:,1]
y_testpred = rf1.predict(Xtest)
newindex = data_test.index
y_testpred=pd.DataFrame(y_testpred,index=newindex)

#SVN
from sklearn.svm import SVC
grid = GridSearchCV(SVC(), param_grid={"C":[0.1, 1, 10], "gamma": [1, 0.1, 0.01]}, cv=4)
grid.fit(Xtrain,Ytrain)
y_pred_svc = grid.predict(Xtest)
y_pred_svc=pd.DataFrame(y_pred_svc,index=newindex)

#adaboost
from sklearn.ensemble import AdaBoostClassifier
bdt = AdaBoostClassifier(DecisionTreeClassifier(max_depth=2, min_samples_split=20, min_samples_leaf=5),
                         algorithm="SAMME",
                         n_estimators=200, learning_rate=0.8)
bdt.fit(Xtrain,Ytrain)
y_pred_ada = bdt.predict(Xtest)
y_pred_ada=pd.DataFrame(y_pred_ada,index=newindex)

#GBDT
from sklearn.ensemble import GradientBoostingClassifier
gbm0 = GradientBoostingClassifier(random_state=10)
gbm0.fit(Xtrain,Ytrain)
y_pred_gbdt = gbm0.predict(Xtest)
y_pred_gbdt=pd.DataFrame(y_pred_gbdt,index=newindex)

result = pd.concat([data_test,y_testpred,y_pred_svc,y_pred_ada,y_pred_gbdt], axis=1)
outputfile = 'output201712.xls'
result.to_excel(outputfile)

由此得到样本数据的四种分类算法的测试结果：

可以看出前面几中较简单的分类算法效果很差，而后面四种算法则效果不错。

除了关心正确率和召回率以外，也可以计算预测为正值的平均收益率，比如说预测结果为正而样本为负，但实际利率为4.6%其实也是很好的结果了，即使为1%也是正不会产生损失的风险，所以计算结果的平均收益率也是一个重要的评估指标。