-
Notifications
You must be signed in to change notification settings - Fork 1
/
visualize_cross_feature.py
362 lines (257 loc) · 14.9 KB
/
visualize_cross_feature.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
import os
import sys
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn import linear_model
from sklearn import svm
from functools import reduce
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.rcParams['axes.unicode_minus'] = False
def get_plot(data_train):
data_train = data_train[data_train['实际功率'] > 0.03 * data_train['实际功率'].max()]
features = data_train.columns.drop(['时间', '实际功率'])
fig = plt.figure(figsize=(20, 10))
for i in range(len(features)):
ax_i = fig.add_subplot(3, 6, i+1)
ax_i.scatter(data_train[features[i]].values, data_train['实际功率'].values, s=2, alpha=0.2)
ax_i.set_xlabel(features[i], fontsize=20)
ax_i.set_ylabel('实际功率')
plt.show()
def visualize_solar_data():
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['hours_float'].values, train_1_new['azimuth'].values, s=2, alpha=0.2)
plt.title('这张图展示了一天中时间与太阳方位的关系')
# 下面
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['hours_float'].values, train_1_new['altitude'].values, s=2, alpha=0.2)
plt.title('这张图展示了一天中时间与太阳仰角的关系')
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['yday'].values, train_1_new['azimuth'].values, s=2, alpha=0.2)
plt.title('这张图展示了一年到头来每天太阳方位的范围变化情况')
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['yday'].values, train_1_new['altitude'].values, s=2, alpha=0.2)
plt.title('这张图展示了一年到头来每天太阳仰角的范围变化情况')
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['yday'].values, train_1_new['distance'].values, s=2, alpha=0.2)
plt.title('这张图展示了一年到头来每天距离太阳距离的变化情况')
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['yday'].values, train_1_new['diameter'].values, s=2, alpha=0.2)
plt.title('这张图展示了一年到头来每天看到的太阳直径的变化情况')
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['distance'].values, train_1_new['diameter'].values, s=2, alpha=0.2)
plt.title('这张图证明了距离太阳的距离和看到的太阳的直径是严格的函数关系,而且是线性关系')
plt.show()
def contouring(series_1, series_2, series_y, grid_num, scope):
feature_1_min = np.floor(series_1.min())
feature_1_max = np.ceil(series_1.max())
feature_2_min = np.floor(series_2.min())
feature_2_max = np.ceil(series_2.max())
the_xx, the_yy = np.meshgrid(np.linspace(feature_1_min, feature_1_max, grid_num),
np.linspace(feature_2_min, feature_2_max, grid_num))
z = []
for j in range(grid_num):
for i in range(grid_num):
x = the_xx[0, :][i]
y = the_yy[:, 0][j]
condition_1 = (np.abs(series_1.values - x) <= scope * (feature_1_max - feature_1_min))
condition_2 = (np.abs(series_2.values - y) <= scope * (feature_2_max - feature_2_min))
tmp = series_y.values[condition_1 & condition_2]
if len(tmp) == 0:
result = np.nan
else:
result = np.nanmean(tmp)
z.append(result)
the_zz = np.array(z).reshape(the_xx.shape)
return the_xx, the_yy, the_zz
def roll_contour(figure, fit_start, fit_end, grid_num):
"""
获取rolling之后得到的等高线
"""
the_vertices_all = np.array([]).reshape(-1, 2)
for i in range(fit_start, fit_end):
vertices = figure.collections[i].get_paths()[0].vertices
vertices[:, 1] = vertices[:, 1] - np.mean(vertices[:, 1])
the_vertices_all = np.concatenate([the_vertices_all, vertices], axis=0)
vertices_all_df = pd.DataFrame(the_vertices_all, columns=['x', 'y'])
vertices_all_df.sort_values(by='x', inplace=True)
gap = np.int(len(vertices_all_df) / grid_num)
point_num = int(np.ceil(len(vertices_all_df) / gap))
agg_index = reduce(lambda x, y: x + y, [[i] * gap for i in range(point_num)])
vertices_all_df['agg_index'] = agg_index[:len(vertices_all_df)]
the_vertices_all_df_agg = vertices_all_df.rolling(window=100, min_periods=0).mean()
return the_vertices_all, the_vertices_all_df_agg
def get_contour(data_train, data_test, grid_num, feature_1, feature_2, scope, line_num=20, fit_start=2, fit_end=13):
xx, yy, zz = contouring(data_train[feature_1], data_train[feature_2], data_train['实际功率'], grid_num, scope)
# 下面三行只是为了获取等高线,不画图
plt.figure()
fig_object = plt.contourf(xx, yy, zz, line_num, alpha=0.75, cmap=plt.cm.hot)
plt.close()
vertices_all, vertices_all_df_agg = roll_contour(fig_object, fit_start, fit_end, grid_num)
# 下面对等高线进行拟合
weights = np.polyfit(vertices_all_df_agg['x'].values, vertices_all_df_agg['y'].values, 10)
y_predict_poly = np.polyval(weights, vertices_all_df_agg['x'].values)
# 下面得到新的特征
y_residual_train_ = data_train[feature_2] - np.polyval(weights, data_train[feature_1].values)
y_residual_test_ = data_test[feature_2] - np.polyval(weights, data_test[feature_1].values)
the_corr = pd.Series(y_residual_train_).corr(data_train['实际功率'])
# 下面开始可视化
fig_in = plt.figure(figsize=(10, 5))
plt.subplot(1, 2, 1)
plt.plot(vertices_all_df_agg['x'].values, vertices_all_df_agg['y'].values, color='blue')
plt.plot(vertices_all_df_agg['x'].values, y_predict_poly, color='green')
plt.scatter(vertices_all[:, 0], vertices_all[:, 1], s=5, alpha=1)
plt.contourf(xx, yy, zz, line_num, alpha=0.75, cmap=plt.cm.hot)
cc = plt.contour(xx, yy, zz, line_num, colors='black')
plt.clabel(cc, inline=True, fontsize=10)
plt.xlabel(feature_1, fontsize=20)
plt.ylabel(feature_2, fontsize=20)
plt.subplot(1, 2, 2)
xx_res, yy_res, zz_res = contouring(data_train[feature_1], pd.Series(y_residual_train_),
data_train['实际功率'], grid_num, scope)
plt.contourf(xx_res, yy_res, zz_res, line_num, alpha=0.75, cmap=plt.cm.hot)
plt.xlabel(feature_1, fontsize=20)
plt.ylabel('%s_rectified' % feature_2, fontsize=20)
plt.show()
# plt.close()
# print('%s:%s' % (feature_1, train_1_new[feature_1].corr(train_1_new['实际功率'])))
# print('%s:%s' % (feature_2, train_1_new[feature_2].corr(train_1_new['实际功率'])))
return fig_in, y_residual_train_, y_residual_test_, the_corr
def feature_cross(data_train, data_test):
train_feature = pd.DataFrame()
test_feature = pd.DataFrame()
columns = ['辐照度', '风速', '风向', '温度', '压强', '湿度', 'azimuth',
'altitude', 'distance', 'hours_float', 'mday', 'yday']
for i in range(len(columns)):
print('begin: %s' % str(i))
feature_1 = columns[i]
for j in range(i + 1, len(columns)):
feature_2 = columns[j]
_, y_residual_train_1, y_residual_test_1, corr_1 = get_contour(data_train, data_test, 100,
feature_1, feature_2, 0.1)
_, y_residual_train_2, y_residual_test_2, corr_2 = get_contour(data_train, data_test, 100,
feature_2, feature_1, 0.1)
col_name_1 = feature_1 + '_' + feature_2
col_name_2 = feature_2 + '_' + feature_1
train_feature[col_name_1] = y_residual_train_1
train_feature[col_name_2] = y_residual_train_2
test_feature[col_name_1] = y_residual_test_1
test_feature[col_name_2] = y_residual_test_2
print('end: %s' % str(i))
data_train_new = pd.concat([data_train, train_feature], axis=1)
data_test_new = pd.concat([data_test, test_feature], axis=1)
return data_train_new, data_test_new
if __name__ == '__main__':
train_1_new = pd.read_csv('./data_new/train_1_new.csv')
train_2_new = pd.read_csv('./data_new/train_2_new.csv')
train_3_new = pd.read_csv('./data_new/train_3_new.csv')
train_4_new = pd.read_csv('./data_new/train_4_new.csv')
test_1_new = pd.read_csv('./data_new/test_1_new.csv')
test_2_new = pd.read_csv('./data_new/test_2_new.csv')
test_3_new = pd.read_csv('./data_new/test_3_new.csv')
test_4_new = pd.read_csv('./data_new/test_4_new.csv')
get_plot(train_1_new)
visualize_solar_data()
x_0, y_residual_train_s, _, _ = get_contour(train_1_new, test_1_new, 100, '温度', '湿度', 0.1)
x_1, _, _, _ = get_contour(train_1_new, test_1_new, 100, '压强', '温度', 0.2)
x_2, _, _, _ = get_contour(train_1_new, test_1_new, 100, '湿度', '辐照度', 0.1, fit_end=15)
x_3, _, _, _ = get_contour(train_1_new, test_1_new, 100, '温度', '辐照度', 0.1, fit_start=3, fit_end=15)
x_4, _, _, _ = get_contour(train_1_new, test_1_new, 100, '风速', '辐照度', 0.1)
x_5, _, _, _ = get_contour(train_1_new, test_1_new, 100, '风向', '辐照度', 0.1)
x_6, _, _, _ = get_contour(train_1_new, test_1_new, 100, '压强', '辐照度', 0.1)
x_7, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'month', '辐照度', 0.1)
x_8, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'yday', '辐照度', 0.1)
x_9, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'mday', '辐照度', 0.1, line_num=50)
x_10, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'azimuth', '辐照度', 0.1) # a good example
x_11, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'altitude', '辐照度', 0.1, fit_end=8)
x_12, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'azimuth', 'altitude', 0.1)
x_13, _, _, _ = get_contour(train_1_new, test_1_new, 100, 'distance', '辐照度', 0.1)
x_14, _, _, _ = get_contour(train_1_new, test_1_new, 100, '温度', 'altitude', 0.1)
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['yday'], train_1_new['辐照度'], s=2, alpha=0.2)
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['yday'], train_1_new['辐照度'], s=2, alpha=0.2)
fig = plt.figure(figsize=(20, 10))
plt.scatter(train_1_new['湿度'], train_1_new['辐照度'], s=2, alpha=0.2)
train_1_new['hours_float'].corr(train_1_new['azimuth'])
train_1_cross, test_1_cross = feature_cross(train_1_new, test_1_new)
train_2_cross, test_2_cross = feature_cross(train_2_new, test_2_new)
train_3_cross, test_3_cross = feature_cross(train_3_new, test_3_new)
train_4_cross, test_4_cross = feature_cross(train_4_new, test_4_new)
train_1_cross.to_csv('./data_cross/train_1_cross.csv', index=False)
train_2_cross.to_csv('./data_cross/train_2_cross.csv', index=False)
train_3_cross.to_csv('./data_cross/train_3_cross.csv', index=False)
train_4_cross.to_csv('./data_cross/train_4_cross.csv', index=False)
test_1_cross.to_csv('./data_cross/test_1_cross.csv', index=False)
test_2_cross.to_csv('./data_cross/test_2_cross.csv', index=False)
test_3_cross.to_csv('./data_cross/test_3_cross.csv', index=False)
test_4_cross.to_csv('./data_cross/test_4_cross.csv', index=False)
cross_col = [x for x in train_1_cross.columns.drop('hours_float') if len(x.split('_')) > 1]
data_train = train_1_cross[train_1_cross['实际功率'] > 0.03 * train_1_cross['实际功率'].max()]
import lightgbm as lgb
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import train_test_split
from sklearn import metrics
from sklearn.preprocessing import PolynomialFeatures
import scipy.stats as stats
data = data_train
lgb_eval = lgb.Dataset(X_test, y_test, reference=train_matrix)
model = lgb.train(params,
train_matrix,
num_round,
valid_sets=lgb_eval,
early_stopping_rounds=early_stopping_rounds,
verbose_eval=1000)
y_pred = model.predict(new_test.fillna(-999))
def naive_lgb(data_x, data_y):
params = {'learning_rate': 0.084, 'max_depth': -1, 'metric': 'mae',
'min_data': 6, 'min_child_weight': 0.001, 'num_leaves': 100,
'objective': 'regression', 'lambda_l2': 1.1, 'nthread': 4,
'early_stopping_rounds': 100, 'verbose_eval': 50,
'num_boost_round': 1000, 'sub_feature': 0.9}
x_train, x_test, y_train, y_test = train_test_split(data_x, data_y, test_size=0.1, random_state=678)
train_data = lgb.Dataset(x_train, y_train)
valid_data = lgb.Dataset(x_test, y_test)
lgb_model = lgb.train(params, train_set=train_data, valid_sets=valid_data)
y_predict_test = lgb_model.predict(x_test)
return metrics.mean_absolute_error(y_test, y_predict_test)
def naive_predict(data, degree=1, x=['辐照度'], y='实发辐照度'):
"""
x里必须要有’辐照度‘
"""
polynomial = PolynomialFeatures(degree=degree) # 二次多项式
x_transformed = polynomial.fit_transform(data[x])
linear_reg_0 = LinearRegression()
linear_reg_0.fit(data[['辐照度']], data['实际功率'])
y_predict_0 = linear_reg_0.predict(data[['辐照度']])
linear_reg_1 = LinearRegression()
linear_reg_1.fit(x_transformed, data[y])
agg_feature = linear_reg_1.predict(x_transformed)
linear_reg_2 = LinearRegression()
linear_reg_2.fit(pd.DataFrame({'agg_feature': agg_feature}), data['实际功率'])
y_predict_2 = linear_reg_2.predict(pd.DataFrame({'agg_feature': agg_feature}))
mae_0 = metrics.mean_absolute_error(y_predict_0, data['实际功率'])
corr = stats.pearsonr(data['辐照度'].values, agg_feature)
mae_2 = metrics.mean_absolute_error(y_predict_2, data['实际功率'])
plt.scatter(data['辐照度'].values, data[y], s=2, alpha=0.3)
plt.scatter(data['辐照度'].values, agg_feature, s=2, alpha=0.3)
print('辐照度 预测 “实际功率”的mae', mae_0)
print('对“实发辐照度”的预测值与“实发辐照度”的相关性', corr)
print('对“实发辐照度”的预测值 预测 “实际功率”的mae', mae_2)
return mae_2
temp = list(map(lambda x: naive_predict(data_train, degree=x), list(range(1, 20))))
plt.plot(range(1, len(temp) + 1), temp)
#
data_x, data_y = data_train.drop(['时间', '实发辐照度', '实际功率', 'diameter'], axis=1), data_train['实发辐照度']
data_train_2 = data_train.copy()
data_train_2['实发辐照度的预测值'] = agg_feature
data_train_2['实发辐照度的预测值_1'] = agg_feature
data_train_2['实发辐照度的预测值_2'] = agg_feature
data_train_2['实发辐照度的预测值_3'] = agg_feature
data_train_2['实发辐照度的预测值_4'] = agg_feature
data_train_2['实发辐照度的预测值_5'] = agg_feature
data_train_2['实发辐照度的预测值_6'] = agg_feature
data_x, data_y = data_train_2.drop(['时间', '实发辐照度', '实际功率', 'diameter'], axis=1), data_train_2['实际功率']
print(naive_lgb(data_x, data_y))