manufacture

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline

train = pd.read_excel(r"C:\Users\Administrator\Desktop\jimsir\天池大赛\manufacture\data_first\train.xlsx")

# 数据概览
train.head()

	ID	TOOL_ID	210X1	210X2	210X3	210X4	210X5	210X6	210X7	210X8	...	750X1444	750X1445	750X1446	750X1447	750X1448	750X1449	750X1450	750X1451	750X1452	Y
0	ID001	N	102.05	0.465	0.27	1.430	67.45	4.620	-0.54	-1.05	...	0.00072	0.00072	25.7	0.00072	0.00072	25.7	0.00072	0.00072	2.400000e+12	2.945079
1	ID002	M	100.95	0.805	0.22	3.477	62.08	3.412	-2.12	1.02	...	0.00072	0.00072	25.5	0.00072	0.00072	25.5	0.00072	0.00072	2.400000e+12	2.955092
2	ID003	L	98.56	0.555	0.24	1.172	56.70	3.080	-2.25	0.88	...	0.00064	0.00064	25.2	0.00064	0.00064	25.2	0.00064	0.00064	2.400000e+12	2.741264
3	ID004	M	100.35	0.901	0.22	3.631	62.25	3.949	-1.98	0.82	...	0.00072	0.00072	26.4	0.00072	0.00072	26.4	0.00072	0.00072	2.400000e+12	2.799336
4	ID005	M	100.25	0.854	0.23	3.429	61.42	3.630	-1.89	1.02	...	0.00072	0.00072	26.4	0.00072	0.00072	26.4	0.00072	0.00072	2.400000e+12	2.692093

5 rows × 8029 columns

train.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 500 entries, 0 to 499
Columns: 8029 entries, ID to Y
dtypes: float64(6279), int64(1739), object(11)
memory usage: 30.6+ MB

train.describe() # 可以观察空值的数量

	210X1	210X2	210X3	210X4	210X5	210X6	210X7	210X8	210X9	210X10	...	750X1444	750X1445	750X1446	750X1447	750X1448	750X1449	750X1450	750X1451	750X1452	Y
count	500.000000	500.000000	500.000000	500.000000	500.000000	500.000000	500.000000	500.000000	500.000000	500.000000	...	486.000000	486.000000	486.000000	486.000000	486.000000	486.000000	486.000000	486.000000	4.860000e+02	500.000000
mean	92.272420	0.511932	0.232420	1.626228	57.126160	3.424444	-0.428580	0.267060	-0.027540	25.854000	...	0.000694	0.000694	26.010700	0.000694	0.000694	26.010700	0.000694	0.000694	2.400000e+12	2.846187
std	27.471434	0.233265	0.104536	1.090267	18.224157	1.148460	1.188766	1.092823	0.120268	1.908069	...	0.000045	0.000045	0.349445	0.000045	0.000045	0.349445	0.000045	0.000045	0.000000e+00	0.200970
min	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	-2.450000	-1.920000	-0.220000	0.000000	...	0.000540	0.000540	24.900000	0.000540	0.000540	24.900000	0.000540	0.000540	2.400000e+12	2.326846
25%	100.100000	0.365750	0.210000	1.019750	56.575000	3.242000	-1.912500	-0.670000	-0.140000	24.400000	...	0.000670	0.000670	25.800000	0.000670	0.000670	25.800000	0.000670	0.000670	2.400000e+12	2.705686
50%	101.300000	0.506000	0.240000	1.293000	62.130000	3.616000	-0.185000	0.400000	-0.015000	26.000000	...	0.000690	0.000690	26.000000	0.000690	0.000690	26.000000	0.000690	0.000690	2.400000e+12	2.840442
75%	101.750000	0.612250	0.270000	1.925750	67.250000	3.864250	0.285000	1.220000	0.080000	27.000000	...	0.000730	0.000730	26.300000	0.000730	0.000730	26.300000	0.000730	0.000730	2.400000e+12	2.980399
max	102.800000	0.970000	0.410000	3.891000	71.980000	5.173000	2.190000	2.230000	0.170000	29.400000	...	0.000880	0.000880	27.100000	0.000880	0.000880	27.100000	0.000880	0.000880	2.400000e+12	3.454556

8 rows × 8018 columns

train_describe = train.describe()
np.sum(train_describe.loc['count'] != 500)

2238

sns.distplot(train_describe.loc['count'])

<matplotlib.axes._subplots.AxesSubplot at 0xf464088>

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-s7qxMm7P-1571391706490)(output_6_1.png)]

train_describe.loc['count'] == 0

210X1       False
210X2       False
210X3       False
210X4       False
210X5       False
            ...  
750X1449    False
750X1450    False
750X1451    False
750X1452    False
Y           False
Name: count, Length: 8018, dtype: bool

print('all missing',np.sum(train_describe.loc['count'] == 0),'\nSeriously missing',np.sum(((train_describe.loc['count'] <=250)&(train_describe.loc['count'] > 0))))

all missing 62 
Seriously missing 136

print(np.sum(((train_describe.loc['count'] == 222))))

135

train_cols_allnull = train_describe.columns[train_describe.loc['count'] == 0]
train_cols_allnull

Index(['220X150', '220X151', '340X12', '340X14', '340X16', '340X32', '340X50',
       '340X52', '340X54', '340X56', '340X58', '340X60', '340X62', '340X64',
       '340X66', '340X68', '340X70', '340X72', '340X74', '340X76', '340X78',
       '340X80', '340X82', '340X84', '340X86', '340X88', '340X90', '340X92',
       '340X94', '340X96', '340X98', '340X100', '340X102', '340X104',
       '340X106', '340X112', '340X114', '340X116', '340X118', '340X120',
       '340X122', '340X131', '340X142', '340X144', '340X146', '340X148',
       '340X150', '340X152', '340X154', '340X156', '340X158', '340X160',
       '340X162', '340X164', '340X166', '340X168', '340X170', '340X172',
       '340X174', '340X176', '340X178', '340X180'],
      dtype='object')

list(set(train.columns) - set(train_cols_allnull))

['330X1069',
 '330X393',
 '750X1150',
 '312X707',
 '344X184',
 '360X643',
 '400X217',
 '750X290',
 '360X185',
 '360X1257',
 '220X376',
   ......
 '750X1261',
 '330X496',
 '311X160',
 '750X616',
 '261X656',
 '312X531',
 '440AX33',
 '750X907',
 ...]

train_step2 = train[list(set(train.columns) - set(train_cols_allnull))] # 非全空的列

tmp = train_describe.columns[train_describe.loc['count'] == 222]
tmp

Index(['312X5', '312X10', '312X15', '312X20', '312X25', '312X30', '312X35',
       '312X40', '312X45', '312X50',
       ...
       '312X744', '312X750', '312X756', '312X762', '312X768', '312X774',
       '312X780', '312X786', '312X792', '312X798'],
      dtype='object', length=135)

null_222 = train_step2[tmp]

null_222

	312X5	312X10	312X15	312X20	312X25	312X30	312X35	312X40	312X45	312X50	...	312X744	312X750	312X756	312X762	312X768	312X774	312X780	312X786	312X792	312X798
0	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
495	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
496	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	107.7	110.0	25.0	110.0	110.1	110.0	101.5	101.6	101.5	110.0
497	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	107.3	110.0	25.0	109.9	110.0	109.9	101.9	101.9	101.7	110.0
498	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	105.4	110.0	25.0	108.8	108.9	108.8	99.9	100.0	99.8	110.0
499	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	106.1	110.0	25.0	110.1	110.1	110.0	101.8	101.9	101.7	110.0

500 rows × 135 columns

np.sum(np.sum(null_222[null_222['312X5']>0].isnull()))

0

null_222[null_222['312X5']>0]

	312X5	312X10	312X15	312X20	312X25	312X30	312X35	312X40	312X45	312X50	...	312X744	312X750	312X756	312X762	312X768	312X774	312X780	312X786	312X792	312X798
7	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	105.9	110.0	25.0	110.1	110.2	110.0	100.4	100.4	100.3	110.0
8	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	107.5	110.0	25.0	110.1	110.2	110.0	100.6	100.6	100.5	110.0
9	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	105.9	110.0	25.0	110.1	110.1	110.0	101.6	101.7	101.6	110.0
10	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	106.0	110.0	25.0	110.1	110.1	110.0	101.4	101.7	101.2	110.0
24	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	104.4	110.0	25.0	110.0	110.1	109.9	98.7	98.9	98.7	110.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
490	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	104.3	110.0	25.0	106.4	106.5	106.3	99.9	99.9	99.8	110.0
496	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	107.7	110.0	25.0	110.0	110.1	110.0	101.5	101.6	101.5	110.0
497	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	107.3	110.0	25.0	109.9	110.0	109.9	101.9	101.9	101.7	110.0
498	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	105.4	110.0	25.0	108.8	108.9	108.8	99.9	100.0	99.8	110.0
499	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0	...	106.1	110.0	25.0	110.1	110.1	110.0	101.8	101.9	101.7	110.0

222 rows × 135 columns

train_step2_null_222 = train_step2.loc[null_222[null_222['312X5']>0].index]
train_step2.shape

(500, 7967)

train_step2_null_222.shape

(222, 7967)

tools = []
for col in train_step2.columns:
    if 'T' in col:
        tools.append(col)

tools

['TOOL (#2)',
 'Tool (#1)',
 'TOOL_ID (#3)',
 'TOOL_ID (#2)',
 'Tool',
 'Tool (#2)',
 'TOOL',
 'TOOL (#1)',
 'TOOL_ID',
 'TOOL_ID (#1)',
 'Tool (#3)']

for tool in tools:
    print('For tool', tool, train_step2[tool].nunique(),train_step2_null_222[tool].nunique())

For tool TOOL (#2) 2 2
For tool Tool (#1) 8 6
For tool TOOL_ID (#3) 2 2
For tool TOOL_ID (#2) 3 3
For tool Tool 2 2
For tool Tool (#2) 3 3
For tool TOOL 3 3
For tool TOOL (#1) 2 1
For tool TOOL_ID 6 6
For tool TOOL_ID (#1) 2 2
For tool Tool (#3) 10 10

train_step2_null_222['Tool (#1)'].unique()

array([ 530, 1113, 1018, 1110, 1245, 2823], dtype=int64)

train_step2['Tool (#1)'].unique()

array([ 329,  215,  530, 1113, 1018, 1110, 1245, 2823], dtype=int64)

np.sum(train_step2['Tool (#1)'] == 215)

141

np.sum(train_step2['Tool (#1)'] == 329)

137

500-278

222

test_A=pd.read_excel(r"C:\Users\Administrator\Desktop\jimsir\天池大赛\manufacture\data_first\testA.xlsx")

test_A['Tool (#1)'].unique()

array([2823,  215,  329, 1110,  530, 1113, 1245,  206, 1018], dtype=int64)

test_A[((test_A['Tool (#1)'] == 215) | (test_A['Tool (#1)'] == 329) |(test_A['Tool (#1)'] == 206) )].shape

(35, 8028)

np.sum(np.sum(test_A[(test_A['Tool (#1)'] == 215) | (test_A['Tool (#1)'] == 329) |(test_A['Tool (#1)'] == 206)][tmp].isnull()))

4725

4275/35

122.14285714285714

np.sum(np.sum(test_A[(test_A['Tool (#1)'] != 215) & (test_A['Tool (#1)'] != 329) &(test_A['Tool (#1)'] != 206)][tmp].isnull()))

0

sns.distplot(train_step2[train_step2['Tool (#1)'] == 215]['Y'])

<matplotlib.axes._subplots.AxesSubplot at 0xf464dc8>

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-a5L44uzJ-1571391706493)(output_34_1.png)]

sns.distplot(train_step2[train_step2['Tool (#1)'] == 329]['Y'])

<matplotlib.axes._subplots.AxesSubplot at 0xf916688>

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-YEQvZCnt-1571391706494)(output_35_1.png)]

sns.distplot(train_step2[((train_step2['Tool (#1)'] != 329) & (train_step2['Tool (#1)'] != 215))]['Y'])

<matplotlib.axes._subplots.AxesSubplot at 0x165d83c8>

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-zhI7VJSm-1571391706494)(output_36_1.png)]

from sklearn.metrics import mean_squared_error

train['Y'].median()

2.8404417463190392

pred = train_step2['Y'].copy()
a=train_step2[train_step2['Tool (#1)'] == 329]
b = train_step2[train_step2['Tool (#1)'] == 215]
c = train_step2[((train_step2['Tool (#1)'] != 215)&(train_step2['Tool (#1)'] != 329))]
pred.loc[a.index] = a['Y'].median()
pred.loc[b.index] = b['Y'].median()
pred.loc[c.index] = c['Y'].median()

pred.value_counts()

2.779034    222
2.891122    141
2.882981    137
Name: Y, dtype: int64

mean_squared_error(train_step2['Y'], pred)

0.03848814140862121

mean_squared_error(y_pred=pred,y_true=train_step2['Y'])

0.03848814140862121

train_step2['Tool (#1)'].value_counts()

215     141
329     137
1110     59
2823     55
1245     46
1018     23
1113     22
530      17
Name: Tool (#1), dtype: int64

pred= train_step2['Y'].copy()
for val in train_step2['Tool (#1)'].unique():
    print(val, train_step2[train_step2['Tool (#1)'] == val].shape[0])
    pred.loc[train_step2[train_step2['Tool (#1)'] == val].index] = train_step2[train_step2['Tool (#1)'] == val]['Y'].mean()

329 137
215 141
530 17
1113 22
1018 23
1110 59
1245 46
2823 55

mean_squared_error(y_pred=pred, y_true=train_step2['Y'])

0.03805754994694719

np.sum(train_describe.loc['min'] == train_describe.loc['max'])

1041

train_cols_allsame = train_describe.columns[train_describe.loc['min'] == train_describe.loc['max']]

train_cols_allsame

Index(['210X17', '210X31', '210X38', '210X79', '210X103', '210X196', '210X197',
       '210X198', '210X227', '220X1',
       ...
       '750X1399', '750X1400', '750X1401', '750X1403', '750X1404', '750X1406',
       '750X1407', '750X1408', '750X1442', '750X1452'],
      dtype='object', length=1041)

test_A_describe = test_A.describe()

np.sum(test_A_describe.loc['min'] == test_A_describe.loc['max'])

1080

test_A_cols_allsame = test_A_describe.columns[test_A_describe.loc['min'] == test_A_describe.loc['max']]

train_minus_test = set(train_cols_allsame) - set(test_A_cols_allsame)
both_allsame = set(train_cols_allsame) - set(train_minus_test)

len(train_cols_allsame)

1041

len(test_A_cols_allsame)

1080

len(train_minus_test)

22

len(both_allsame)

1019

len(set(train_step2.columns))

7967

cols = set(train_step2.columns) - both_allsame
train_step3 = train_step2[list(cols)]

tools

['TOOL (#2)',
 'Tool (#1)',
 'TOOL_ID (#3)',
 'TOOL_ID (#2)',
 'Tool',
 'Tool (#2)',
 'TOOL',
 'TOOL (#1)',
 'TOOL_ID',
 'TOOL_ID (#1)',
 'Tool (#3)']

train_tool = train[tools]

from sklearn.preprocessing import LabelEncoder

labelEncoder = LabelEncoder()
a_list = ['a'] * 2 + ['b']*3 + ['c','b']
print(a_list)
labelEncoder.fit_transform(a_list)

['a', 'a', 'b', 'b', 'b', 'c', 'b']





array([0, 0, 1, 1, 1, 2, 1], dtype=int64)

from sklearn.preprocessing import LabelEncoder

for tool in tools:
    lb = LabelEncoder()
    lb.fit(train_tool[tool])
    train_tool.loc[:,tool] = lb.transform(train_tool[tool])

D:\Anaconda3\envs\python3\lib\site-packages\pandas\core\indexing.py:494: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  self.obj[item] = s
D:\Anaconda3\envs\python3\lib\site-packages\pandas\core\indexing.py:576: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  self.obj[item_labels[indexer[info_axis]]] = value

train.shape

(500, 8029)

train_describe.shape

(8, 8018)

#结合交叉验证，构建随机森林模型
from sklearn import model_selection
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
rf = RandomForestRegressor(n_estimators=200,min_samples_leaf=6,min_samples_split=6)
val = model_selection.cross_val_score(rf,train_tool,train['Y'].values,cv=10,scoring='neg_mean_squared_error')

# 评估
print(val)
print(np.mean(val))

[-0.03544356 -0.02205482 -0.03145397 -0.05643132 -0.02725092 -0.04143731
 -0.02598229 -0.03891663 -0.03350271 -0.03271177]
-0.03451853149088352

train_tool_dummy = train_tool.copy()
for tool in tools:
    train_tool_dummy = pd.get_dummies(data=train_tool_dummy, columns=[tool])

val = model_selection.cross_val_score(rf, train_tool_dummy,train['Y'].values,cv=10,scoring='neg_mean_squared_error')

np.mean(val)

-0.033536821368123956

# 利用train_test_split重新训练模型
from sklearn.model_selection import train_test_split
train_X, test_X, train_Y, test_Y = train_test_split(train_tool_dummy,train['Y'].values, test_size=0.3)
rf=RandomForestRegressor(n_estimators=200,min_samples_leaf=6,min_samples_split=6)
rf.fit(train_X, train_Y)

RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=None,
                      max_features='auto', max_leaf_nodes=None,
                      min_impurity_decrease=0.0, min_impurity_split=None,
                      min_samples_leaf=6, min_samples_split=6,
                      min_weight_fraction_leaf=0.0, n_estimators=200,
                      n_jobs=None, oob_score=False, random_state=None,
                      verbose=0, warm_start=False)

#评估
pred = rf.predict(test_X)
mean_squared_error(y_pred=pred,y_true=test_Y)

0.03561266686805807

compare = pd.DataFrame({'y_pred':pred,'y_true':test_Y})
compare['diff'] = compare['y_true'] - compare['y_pred']
compare.sort_values('diff')

	y_pred	y_true	diff
53	2.941684	2.485652	-0.456033
139	2.984927	2.569413	-0.415514
145	2.728678	2.350377	-0.378301
84	2.860819	2.491770	-0.369049
121	2.757900	2.397770	-0.360130
...	...	...	...
113	2.836899	3.230493	0.393593
45	2.828529	3.232169	0.403640
120	2.902924	3.321433	0.418509
20	2.787306	3.270821	0.483515
125	2.745893	3.238402	0.492509

150 rows × 3 columns

train_X, test_X, train_Y, test_Y = train_test_split(train_tool_dummy,train['Y'].values, test_size=0.3)
rf=RandomForestRegressor(n_estimators=200,min_samples_leaf=6,min_samples_split=6)
rf.fit(train_X, train_Y)

RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=None,
                      max_features='auto', max_leaf_nodes=None,
                      min_impurity_decrease=0.0, min_impurity_split=None,
                      min_samples_leaf=6, min_samples_split=6,
                      min_weight_fraction_leaf=0.0, n_estimators=200,
                      n_jobs=None, oob_score=False, random_state=None,
                      verbose=0, warm_start=False)

for i in range(len(train_tool.columns)):
    print(train_tool.columns[i],rf.feature_importances_[i])

TOOL (#2) 0.0
Tool (#1) 0.0
TOOL_ID (#3) 0.04977919950874401
TOOL_ID (#2) 0.034050966856127914
Tool 0.0006262455932305422
Tool (#2) 0.0
TOOL 0.00987126347542283
TOOL (#1) 0.002039560967227642
TOOL_ID 0.008330845725601452
TOOL_ID (#1) 0.01961358235658726
Tool (#3) 0.014775338277854095

train_corr = train_step3.corr()

strong_linear_fea=[]
for col in train_corr.columns:
    if np.abs(train_corr.loc['Y',col]) >=0.1 and col != 'Y':
        strong_linear_fea.append(col)
strong_linear_fea

['330X1069',
 '330X393',
 '750X1150',
 '360X185',
 '261X352',
 '330X300',
 '520X260',
 '310X176',
 '750X197',
  ...
 '360X726',
 '420X90',
 '750X153',
 '360X1059',
 '330X557',
 '261X241',
 '330X1072',
 '210X152',
 '750X1389',
 '330X853',
 '750X164',
 '330X599',
 '750X869',
 ...]

train_tool_linear = train_tool_dummy.copy()
for col in strong_linear_fea:
    train_tool_linear[col] = train_step3[col]

train_tool_linear.fillna(-100,inplace=True)

rf=RandomForestRegressor(n_estimators=200,min_samples_leaf=5,min_samples_split=5)
val = model_selection.cross_val_score(rf, train_tool_dummy,train['Y'].values,cv=10,scoring='neg_mean_squared_error')

np.mean(val)

-0.0335044546686339

from sklearn.svm import SVR
train_tool_linear = train_tool_dummy.copy()
for col in strong_linear_fea:
    if np.max(train_step3[col]) == np.min(train_step3[col]):
        continue
    train_tool_linear[col] = (train_step3[col] - np.min(train_step3[col]))/(np.max(train_step3[col])-np.min(train_step3[col]))
    
sv = SVR(C=2.5)
train_tool_linear.fillna(-1,inplace=True)
val = model_selection.cross_val_score(sv,train_tool_linear,train['Y'].values,cv=10,scoring='neg_mean_squared_error')

#评估
np.mean(val)

D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)
D:\Anaconda3\envs\python3\lib\site-packages\sklearn\svm\base.py:193: FutureWarning: The default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.
  "avoid this warning.", FutureWarning)

-0.030984216663715352