-
Notifications
You must be signed in to change notification settings - Fork 0
/
model_modules.py
372 lines (363 loc) · 20.2 KB
/
model_modules.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
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
# packages
import pandas as pd
import numpy as np
import random
import os
from sklearn.metrics import mean_squared_error
from sklearn.preprocessing import MinMaxScaler
from scipy.stats import spearmanr
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, LSTM
from tensorflow import keras
import tensorflow as tf
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.metrics import r2_score
import sklearn
from pathlib import Path
import time
import os
import matplotlib.dates as mdates
import matplotlib.gridspec as gridspec
import matplotlib.font_manager as fm
import matplotlib
import pydot
from keras.utils import plot_model
import matplotlib.dates as mdates
from matplotlib.dates import DayLocator, HourLocator, DateFormatter, drange
import statsmodels
from statsmodels.graphics.tsaplots import plot_acf
from statsmodels.tsa.stattools import acf
from statsmodels.stats.diagnostic import acorr_breusch_godfrey
import statsmodels.tsa.api as tsa
from scipy.stats import probplot, moment
import statsmodels.api as sm
from arch.unitroot import PhillipsPerron, KPSS
# for general tuning
class prepare_LSTM:
def __init__(self, raw_data_loc):
self.raw_data_loc= raw_data_loc
def data_cleaning(self):
xformatter = mdates.DateFormatter('%d.%b.%Y')
data = pd.read_csv(self.raw_data_loc)
# cleaning trading hours - normal days
data["Local time"]= data["Local time"].astype(str).str[0:16]
sel= []
for i in range(len(data)):
time= int(data['Local time'][i][11:13] + data['Local time'][i][14:16])
if 900 <= time <= 1700:
sel.append(i)
data= data.iloc[sel]
data= data.reset_index(drop=True)
# cleaning trading hours - special days
spec_days = []
for i in range(len(data)):
if data['Local time'][i][0:5] == '31.12' or data['Local time'][i][0:5] == '24.12':
time= int(data['Local time'][i][11:13] + data['Local time'][i][14:16])
if 900 <= time <= 1300:
spec_days.append(i)
else:
spec_days.append(i)
data= data.iloc[spec_days]
data= data.reset_index(drop=True)
# cleaning days
data['date'] = pd.to_datetime(data["Local time"], format= '%d.%m.%Y %H:%M').dt.strftime('%d.%m.%Y')
data_no_vol= data.groupby(['date'], as_index=False)['Volume'].sum()
no_vol= data_no_vol.date[data_no_vol.Volume == 0].to_list()
sel_= []
for i in range(len(data)):
if data.date[i] in no_vol:
pass
else:
sel_.append(i)
data= data.iloc[sel_]
data= data.reset_index(drop=True)
data= data.set_index(["Local time"])
# deleting 2015-05-01
data= data[(data.index < "01.05.2015 09:00") | (data.index > "01.05.2015 17:00")]
# price transformation
data["Price"]= (data.Low + data.High) / 2.0
data= data[["Price"]]
return data
def data_LSTM(self):
scaler = MinMaxScaler(feature_range=(0, 1))
data= prepare_LSTM(self.raw_data_loc).data_cleaning()
dataset = scaler.fit_transform(data.values.astype('float32'))
train= dataset[0:len(data[:'31.12.2019 13:00']),:]
return dataset, train, data, scaler
class split_LSTM:
def __init__(self, dataset, window_size):
self.dataset= dataset
self.window_size= window_size
def create_dataset(self):
dataX, dataY = [], []
for i in range(len(self.dataset)-self.window_size):
a = self.dataset[i:(i+self.window_size), 0]
dataX.append(a)
dataY.append(self.dataset[i + self.window_size, 0])
return np.array(dataX), np.array(dataY)
def model_SPLITS(self, data, train):
validation= self.dataset[(len(data[:'31.12.2019 13:00']) - self.window_size):len(data[:'31.08.2020 17:00']),:]
trainX, y_train = split_LSTM(train, self.window_size).create_dataset()
valX, y_val = split_LSTM(validation, self.window_size).create_dataset()
X_train= np.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
X_val = np.reshape(valX, (valX.shape[0], 1, valX.shape[1]))
return y_train, y_val, X_train, X_val, trainX, valX, validation
class render_result:
def __init__(self, data, save_loc):
self.data= data
self.save_loc= save_loc
def create_table(self, col_width=2.0, row_height=0.6, font_size=16,
header_color='#760002', row_colors=['#f1f1f2', 'w'], edge_color='w',
bbox=[0, 0, 1, 1], header_columns=0,
ax=None, **kwargs):
colors = self.data.shape[0] * [self.data.shape[1] * ["#e9f1f7"]]
if ax is None:
size = (np.array(self.data.shape[::-1]) + np.array([0, 1])) * np.array([col_width, row_height])
fig, ax = plt.subplots(figsize= size)
ax.axis('off')
mpl_table = ax.table(cellText= self.data.values, bbox=bbox, colLabels= self.data.columns, **kwargs, cellLoc= "center", cellColours= colors)
mpl_table.auto_set_font_size(False)
mpl_table.set_fontsize(font_size)
for k, cell in mpl_table._cells.items():
cell.set_edgecolor(edge_color)
if k[0] == 0 or k[1] < header_columns:
cell.set_text_props(weight='bold', color='w', multialignment= 'center', fontname="Times New Roman")
cell.set_facecolor(header_color)
else:
cell.set_text_props(color='black', horizontalalignment= 'center', fontname="Times New Roman")
return ax.get_figure(), ax
set_align_for_column(mpl_table, align="center")
class run_LSTM:
def __init__(self, window_size, neurons= 54, batch= 76):
self.window_size= window_size
self.neurons= neurons
self.batch= batch
def model_BASE(self, scaler, i, y_train, y_val, X_train, X_val, trainX, valX, validation):
result = pd.DataFrame(columns= ["Results RMSE", "Train", "Validation", "Windows size", "Neurons", "Batch size"]) # to store values
tf.random.set_seed(1234)
rnn = Sequential()
rnn.add(LSTM(units= self.neurons, activation='tanh', input_shape= (1, self.window_size), name='LSTM'))
rnn.add(Dense(1))
rnn.compile(loss= 'mean_squared_error',
optimizer='rmsprop')
early_stopping = EarlyStopping(monitor= 'val_loss',
patience= 20,
restore_best_weights= True)
lstm_training = rnn.fit(X_train,
y_train,
epochs= 150,
batch_size= self.batch,
shuffle= True,
validation_data= (X_val, y_val),
callbacks= [early_stopping],
verbose= 0)
train_predict_scaled = rnn.predict(X_train)
val_predict_scaled = rnn.predict(X_val)
train_pred_ = scaler.inverse_transform(train_predict_scaled.reshape(train_predict_scaled.shape[0], train_predict_scaled.shape[1]))
trainY = scaler.inverse_transform([y_train])
val_pred_ = scaler.inverse_transform(val_predict_scaled.reshape(val_predict_scaled.shape[0], val_predict_scaled.shape[1]))
valY = scaler.inverse_transform([y_val])
train_rmse = np.sqrt(mean_squared_error(train_pred_.reshape(-1,), trainY.reshape(-1,)))
validation_rmse = np.sqrt(mean_squared_error(val_pred_.reshape(-1,), valY.reshape(-1,)))
return [str("Model " + str(i)), str(f'{train_rmse:.3f}'), str(f'{validation_rmse:.3f}'), str(self.window_size), str(self.neurons), str(self.batch)]
# for final model, includes test sample
class split_final_LSTM:
def __init__(self, dataset, window_size):
self.dataset= dataset
self.window_size= window_size
def create_dataset(self):
dataX, dataY = [], []
for i in range(len(self.dataset)-self.window_size):
a = self.dataset[i:(i+self.window_size), 0]
dataX.append(a)
dataY.append(self.dataset[i + self.window_size, 0])
return np.array(dataX), np.array(dataY)
def model_final_SPLITS(self, data, train):
validation= self.dataset[(len(data[:'31.12.2019 13:00']) - self.window_size):len(data[:'31.08.2020 17:00']),:]
test= self.dataset[(len(data[:'31.08.2020 17:00']) - self.window_size):,:] #
trainX, y_train = split_LSTM(train, self.window_size).create_dataset()
valX, y_val = split_LSTM(validation, self.window_size).create_dataset()
testX, y_test = split_LSTM(test, self.window_size).create_dataset() #
X_train= np.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
X_val = np.reshape(valX, (valX.shape[0], 1, valX.shape[1]))
X_test = np.reshape(testX, (testX.shape[0], 1, testX.shape[1]))
return y_train, y_val, y_test, X_train, X_val, X_test, trainX, valX, testX, validation, test
class run_final_LSTM:
def __init__(self, window_size= 15, neurons= 38, batch= 84):
self.window_size= window_size
self.neurons= neurons
self.batch= batch
def MAPE(self, Y_Predicted, Y_actual):
mape = np.mean(np.abs((Y_actual - Y_Predicted)/Y_actual))*100
return mape
def model_final_BASE(self, scaler, i, y_train, y_val, y_test, X_train, X_val, X_test, trainX, valX, testX, validation, test):
result = pd.DataFrame(columns= ["Results RMSE", "Train", "Validation", "Windows size", "Neurons", "Batch size"]) # to store values
tf.random.set_seed(1234)
rnn = Sequential()
rnn.add(LSTM(units= self.neurons, activation='relu', input_shape= (1, self.window_size), name='LSTM'))
rnn.add(Dense(1))
rnn.compile(loss= 'mean_squared_error',
optimizer='rmsprop')
early_stopping = EarlyStopping(monitor= 'val_loss',
patience= 20,
restore_best_weights= True)
lstm_training = rnn.fit(X_train,
y_train,
epochs= 150,
batch_size= self.batch,
shuffle= True,
validation_data= (X_val, y_val),
callbacks= [early_stopping],
verbose= 0)
train_predict_scaled = rnn.predict(X_train)
val_predict_scaled = rnn.predict(X_val)
test_predict_scaled = rnn.predict(X_test)
train_pred_ = scaler.inverse_transform(train_predict_scaled.reshape(train_predict_scaled.shape[0], train_predict_scaled.shape[1]))
trainY = scaler.inverse_transform([y_train])
val_pred_ = scaler.inverse_transform(val_predict_scaled.reshape(val_predict_scaled.shape[0], val_predict_scaled.shape[1]))
valY = scaler.inverse_transform([y_val])
test_pred_ = scaler.inverse_transform(test_predict_scaled.reshape(test_predict_scaled.shape[0], test_predict_scaled.shape[1]))
testY = scaler.inverse_transform([y_test])
train_rmse = np.sqrt(mean_squared_error(train_pred_.reshape(-1,), trainY.reshape(-1,)))
validation_rmse = np.sqrt(mean_squared_error(val_pred_.reshape(-1,), valY.reshape(-1,)))
test_rmse = np.sqrt(mean_squared_error(test_pred_.reshape(-1,), testY.reshape(-1,)))
train_ic = spearmanr(y_train, train_predict_scaled)[0]
val_ic = spearmanr(y_val, val_predict_scaled)[0]
test_ic = spearmanr(y_test, test_predict_scaled)[0]
train_mape = run_final_LSTM().MAPE(train_pred_.reshape(-1,), trainY.reshape(-1,))
validation_mape = run_final_LSTM().MAPE(val_pred_.reshape(-1,), valY.reshape(-1,))
test_mape = run_final_LSTM().MAPE(test_pred_.reshape(-1,), testY.reshape(-1,))
return lstm_training, testY, trainY, valY, test_pred_, train_pred_, val_pred_, train_rmse, validation_rmse, test_rmse, train_mape, validation_mape, test_mape, train_ic, val_ic, test_ic
# for graphs of the final model
class graphs_final_model(run_final_LSTM):
def __init__(self, window_size):
super().__init__(window_size)
def create_graphs(self, data, lstm_training, testY, trainY, valY, test_pred_, train_pred_, val_pred_):
xformatter = mdates.DateFormatter('%d.%b.%Y')
with sns.axes_style('whitegrid'):
fig, ax = plt.subplots(figsize=(14, 6))
plt.xticks(fontsize=16, fontname="Times New Roman")
plt.xlabel("Epochs", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
loss_history = pd.DataFrame(lstm_training.history).pow(.5)
loss_history.index += 1
best_rmse = loss_history.val_loss.min()
best_epoch = loss_history.val_loss.idxmin()
title = f'Best validation RMSE: {best_rmse:.4%})'
loss_history.columns= ['Training RMSE', 'Validation RMSE']
loss_history.rolling(5).mean().plot(logy=True, lw= 3, ax=ax, color= ['darkgreen', 'darkorange'])
ax.set_title(title, fontsize=18, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax.axvline(best_epoch, ls='--', lw=3, c='red')
sns.despine()
ax.legend(prop={'size': 16, "family":"Times New Roman"})
ax.set_xticks(np.arange(8, len(loss_history), 10).tolist() + [best_epoch, len(loss_history)])
ax.set_facecolor('#E9F1F7')
fig.tight_layout()
plt.savefig('figures/Model_evaluation.png')
with sns.axes_style('whitegrid'):
fig, axes = plt.subplots(nrows=1, ncols=1, figsize=(16, 10))
gs = gridspec.GridSpec(1, 1, wspace=0.1, hspace=0.1)
ax1 = plt.subplot(gs[0])
plt.xticks(fontsize=16, fontname="Times New Roman")
plt.yticks(fontsize=16, fontname="Times New Roman")
Real= pd.Series(testY.reshape(-1,), pd.to_datetime(data[len(data[:'31.08.2020 17:00']):].index, format= '%d.%m.%Y %H:%M'))
Real.plot(c= 'darkred', legend= True, linewidth= 1, ax= ax1)
Predicted= pd.Series(test_pred_.reshape(-1,), index= pd.to_datetime(data[len(data[:'31.08.2020 17:00']):].index, format= '%d.%m.%Y %H:%M'))
Predicted.plot(c= 'blue', legend= True, linewidth= 2, ax= ax1)
ax1.set_xlabel("Date", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax1.set_ylabel("Mid price", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax1.legend(['Real', 'Prediction on test sample'], prop={'size': 16, "family":"Times New Roman"})
ax1.fmt_xdata = DateFormatter('%d.%b.%Y')
ax1.xaxis.set_major_formatter(xformatter)
ax1.set_facecolor('#E9F1F7')
plt.setp(ax1.xaxis.get_majorticklabels(), rotation= 0, ha= 'center')
plt.savefig('figures/Model_predict_test.png')
with sns.axes_style('whitegrid'):
fig, axes = plt.subplots(nrows=2, ncols=1, figsize=(16, 12))
gs = gridspec.GridSpec(2, 1, wspace=0.4, hspace=0.3)
ax1 = plt.subplot(gs[0])
plt.xticks(fontsize=16, fontname="Times New Roman")
plt.yticks(fontsize=16, fontname="Times New Roman")
ax2 = plt.subplot(gs[1])
plt.xticks(fontsize=16, fontname="Times New Roman")
plt.yticks(fontsize=16, fontname="Times New Roman")
Real1= pd.Series(trainY.reshape(-1,), pd.to_datetime(data[self.window_size:len(data[:'31.12.2019 13:00'])].index, format= '%d.%m.%Y %H:%M'))
Real1.plot(c= '#760002', legend= True, linewidth= 1, ax= ax1)
Predicted1= pd.Series(train_pred_.reshape(-1,), index= pd.to_datetime(data[self.window_size:len(data[:'31.12.2019 13:00'])].index, format= '%d.%m.%Y %H:%M'))
Predicted1.plot(c= 'darkgreen', legend= True, linewidth= 2, ax= ax1)
ax1.set_xlabel("Date", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax1.set_ylabel("Mid price", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax1.legend(['Real', 'Prediction on training sample'], prop={'size': 16, "family":"Times New Roman"})
ax1.fmt_xdata = DateFormatter('%d.%b.%Y')
ax1.xaxis.set_major_formatter(xformatter)
ax1.set_facecolor('#E9F1F7')
plt.setp(ax1.xaxis.get_majorticklabels(), rotation= 0, ha= 'center')
Real2= pd.Series(valY.reshape(-1,), pd.to_datetime(data[len(data[:'31.12.2019 13:00']):len(data[:'31.08.2020 17:00'])].index, format= '%d.%m.%Y %H:%M'))
Real2.plot(c= '#760002', legend= True, linewidth= 1, ax= ax2)
Predicted2= pd.Series(val_pred_.reshape(-1,), index= pd.to_datetime(data[len(data[:'31.12.2019 13:00']):len(data[:'31.08.2020 17:00'])].index, format= '%d.%m.%Y %H:%M'))
Predicted2.plot(c= 'darkorange', legend= True, linewidth= 2, ax= ax2)
ax2.set_xlabel("Date", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax2.set_ylabel("Mid price", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax2.legend(['Real', 'Prediction on validation sample'], prop={'size': 16, "family":"Times New Roman"})
ax2.fmt_xdata = DateFormatter('%d.%b.%Y')
ax2.xaxis.set_major_formatter(xformatter)
ax2.set_facecolor('#E9F1F7')
plt.setp(ax2.xaxis.get_majorticklabels(), rotation= 0, ha= 'center')
plt.savefig('figures/Model_predict_val_train.png')
plt.close()
# for econometric testing
class econ_test:
def __init__(self, raw_data_loc, window_size):
self.raw_data_loc= raw_data_loc
self.window_size= window_size
def data_prep(self):
data= prepare_LSTM(self.raw_data_loc).data_cleaning()
components = tsa.seasonal_decompose(data, model='additive', period= self.window_size)
ts = (data
.assign(Trend= components.trend)
.assign(Seasonality= components.seasonal)
.assign(Residual= components.resid))
ts.index= pd.to_datetime(data.index, format= '%d.%m.%Y %H:%M')
X= data.values
return data, ts, components, X
def print_econ_mod(self, model, name):
plt.rc('figure', figsize=(6, 3))
plt.text(0, 0, str(model.summary()), {'fontsize': 16, 'fontname':"Times New Roman"})
plt.axis('off')
plt.tight_layout()
plt.savefig('figures/' + name + '.png')
plt.close()
def pp_test(self):
data, ts, components, X= econ_test(self.raw_data_loc, self.window_size).data_prep()
pp = PhillipsPerron(X)
pp.lags = self.window_size
pp.trend = 'ct'
econ_test(self.raw_data_loc, self.window_size).print_econ_mod(pp, 'pp_test')
def kpss_test(self):
data, ts, components, X= econ_test(self.raw_data_loc, self.window_size).data_prep()
kpss = KPSS(X)
kpss.lags= self.window_size
kpss.trend = 'ct'
econ_test(self.raw_data_loc, self.window_size).print_econ_mod(kpss, 'kpss_test')
def autocorr_graph(self, name):
data, ts, components, X= econ_test(self.raw_data_loc, self.window_size).data_prep()
data_series= pd.Series(data.Price, index= data.index)
data_log = np.log(data_series)
data_log_diff = data_log.diff().dropna()
data_log_diff.index= pd.to_datetime(data_log_diff.index, format= '%d.%m.%Y %H:%M')
xformatter = mdates.DateFormatter('%d.%b.%Y')
with sns.axes_style('whitegrid'):
fig, axes = plt.subplots(nrows=1, ncols=1, figsize=(16, 6))
gs = gridspec.GridSpec(1, 1, wspace=0.4, hspace=0.3)
ax1 = plt.subplot(gs[0])
plt.xticks(fontsize=16, fontname="Times New Roman")
plt.yticks(fontsize=16, fontname="Times New Roman")
ax1.set_xticks(np.arange(0, 289, 36).tolist())
plot_acf(data_log_diff, lags= self.window_size, zero=False, ax= ax1, c= '#760002')
ax1.set_title("Autocorrelation", fontsize=18, fontname="Times New Roman", fontdict= dict(weight='bold'))
ax1.set_xlabel("Lags", fontsize=16, fontname="Times New Roman", fontdict=dict(weight='bold'))
ax1.set_facecolor('#E9F1F7')
ax1.set_xticks(np.arange(1, self.window_size+1, 1).tolist())
plt.setp(ax1.xaxis.get_majorticklabels(), rotation= 0, ha= 'center')
fig.savefig('figures/'+ name + '.png')