from sklearn.feature_selection import SelectFromModel
from sklearn.ensemble import RandomForestRegressor
Def selectFromModel():
stock =pd.read_csv('stock.csv',encoding='GBK')
y = stock['涨跌幅’]
features = stock.loc[:,'价格':'流通市值']
X = features.values
X_train, X_test, y_train,y_test = train_test_split(X, y, random_state=62)
    #预处理
scaler = StandardScaler()
scaler.fit(X_train)
X_train_scaled =scaler.transform(X_train)
X_test_scaled =scaler.transform(X_test)
sfm = SelectFromModel(estimator=RandomForestRegressor(n_estimators=100,random_state=38),threshold='median')
sfm.fit(X_train_scaled,y_train)
X_train_sfm =sfm.transform(X_train_scaled)                   
print('经过随机森林模型特征选择后的的数据形态:{}'.format(X_train_sfm.shape))
mask = sfm.get_support()
print(mask)

经过随机森林模型进行特征后的的数据形态:(306, 8)
[FalseTrue False False False False True False False True True False True True True True]

#使用随机森林特征选择后数据集训练神经网络
       X_test_sfm =sfm.transform(X_test_scaled)
       mlp_sfm =MLPRegressor(random_state=62,hidden_layer_sizes=[100,200,100],alpha=0.1)
       mlp_sfm.fit(X_train_sfm,y_train)
print('经过随机森林模型特征选择后训练集得分:{:.2%}'.format(mlp_sfm.score(X_train_sfm,y_train)))
print('经过随机森林模型特征选择后测试集得分:{:.2%}'.format(mlp_sfm.score(X_test_sfm,y_test)))

经过随机森林模型特征选择后训练集得分:96.75%
经过随机森林模型特征选择后测试集得分:97.31%

from sklearn.feature_selection import RFE
def elimination():
    stock =pd.read_csv('stock.csv',encoding='GBK')
    y = stock['涨跌幅']
    features = stock.loc[:,'价格':'流通市值']
    X = features.values
    X_train, X_test, y_train,y_test = train_test_split(X, y, random_state=62)
#预处理
    scaler = StandardScaler()
    scaler.fit(X_train)
    X_train_scaled =scaler.transform(X_train)
    X_test_scaled =scaler.transform(X_test)
    rfe = RFE(RandomForestRegressor(n_estimators=100,random_state=38),n_features_to_select=8)
   rfe.fit(X_train_scaled,y_train)
    X_train_rfe =rfe.transform(X_train_scaled)
print('经过随机森林模型进行迭代特征选择后的的数据形态:{}'.format(X_train_rfe.shape))
mask = rfe.get_support()
print(mask)
print(mask)
#用图像表示特征选择结果
  plt.matshow(mask.reshape(1,-1),cmap=plt.cm.cool)
  plt.xlabel(u"特征选择")
  plt.rcParams['font.sans-serif']=['SimHei’]
  plt.rcParams['axes.unicode_minus']=False
plt.show()

经过随机森林模型进行迭代特征选择后的的数据形态:(306, 8)
[FalseTrue True False False False False False False True True True True True TrueFalse]

 使用随机森林迭代特征选择后数据集训练随机森林
       X_test_rfe =rfe.transform(X_test_scaled)
       mlp_rfe =MLPRegressor(random_state=62,hidden_layer_sizes=[100,200,100],alpha=0.1)
       mlp_rfe.fit(X_train_rfe,y_train)
print('经过随机森林迭代特征选择后训练集得分:{:.2%}'.format(mlp_rfe.score(X_train_rfe,y_train)))
print('经过随机森林迭代特征选择后测试集得分:{:.2%}'.format(mlp_rfe.score(X_test_rfe,y_test)))

经过随机森林迭代特征选择后训练集得分:96.64%
经过随机森林迭代特征选择后测试集得分:96.66%

def summary():
stock =pd.read_csv('stock.csv',encoding='GBK')
y = stock['涨跌幅']
features = stock.loc[:,'价格':'流通市值']
X = features.values
clf_sp =SelectPercentile(percentile=50)
clf_sp.fit(X,y)
sp_masks =clf_sp.get_support()
clf_sfm =SelectFromModel(estimator=RandomForestRegressor(n_estimators=100,random_state=38),threshold='median')
clf_sfm.fit(X,y)
sfm_masks =clf_sfm.get_support()
clf_rfe =RFE(RandomForestRegressor(n_estimators=100,random_state=38),n_features_to_select=8)
clf_rfe.fit(X,y)
rfe_masks =clf_rfe.get_support()
merage_masks =[]
for sp_mask,sfm_mask,rfe_maskin zip(sp_masks,sfm_masks,rfe_masks):
merage = sp_maskand sfm_mask and rfe_mask
merage_masks.append(merage)    
i = 0
New_X =np.empty([X.shape[0],1])
for merage_mask in(merage_masks):
if merage_mask:
New_X = np.column_stack((X[:,i],New_X))
i = i+1
New_X=np.delete(New_X,0,1)
X_train, X_test,y_train, y_test = train_test_split(New_X, y, random_state=88)
scaler =StandardScaler()
scaler.fit(X_train)
X_train_scaled =scaler.transform(X_train)
X_test_scaled =scaler.transform(X_test)
mlp_merage =MLPRegressor(random_state=62,hidden_layer_sizes=[100,200,100],alpha=0.1)
mlp_merage.fit(X_train,y_train)
print('训练集得分:{:.2%}'.format(mlp_merage.score(X_train,y_train)))
print('测试集得分:{:.2%}'.format(mlp_merage.score(X_test,y_test)))

训练集得分:36.97%
测试集得分:51.17%