-
Notifications
You must be signed in to change notification settings - Fork 2
/
tests.py
400 lines (304 loc) · 14.3 KB
/
tests.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
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
"""
# -- --------------------------------------------------------------------------------------------------- -- #
# -- project: Applications of Genetic Methods for Feature Engineering and Hyperparameter Optimization -- #
# -- -------- for Neural Networks. -- #
# -- script: main.py : python script with the main functionality of the project -- #
# -- author: IFFranciscoME - [email protected] -- #
# -- license: GPL-3.0 License -- #
# -- repository: https://github.com/IFFranciscoME/GeneticMethods -- #
# -- --------------------------------------------------------------------------------------------------- -- #
"""
# -- Load other scripts
import functions as fn
import data as dt
import visualizations as vs
# -- Load packages for this script
import numpy as np
import pandas as pd
import ccxt
import seaborn as sns
import matplotlib.pyplot as plt
# Minor adjustments for pandas visualizations
pd.set_option('display.max_rows', None) # unlimit rows
pd.set_option('display.max_columns', None) # unlimit cols
pd.set_option('display.width', None) # unlimit width display
pd.set_option('display.expand_frame_repr', False) # expand cols
pd.options.mode.chained_assignment = None # no index warning
# --------------------------------------------------------------------------------------------- Notebook -- #
# Use ccxt library
# list of available exchanges in ccxt
exchanges = ccxt.exchanges
# Use previously constructed function to fetch historical OHLCV data with CCXT - Binance Public API
# help(dt.ini_binance)
# help(dt.massive_ohlcv)
# --------------------------------------------------------------------------------------------- Notebook -- #
## Get historical OHLC Prices
# Get historical data (previously downloaded - check data.py)
df_data = dt.df_prices
# First 5 elements
# df_data.head(5)
# Last 5 elements
df_data.tail(5)
# General description
data_profile = fn.data_profile(p_data=df_data.copy(), p_type='ohlc', p_mult=1)
print(data_profile)
# --------------------------------------------------------------------------------------------- Notebook -- #
## Visualize data with OHLC Candlestick plot made with plotly
# Plot Financial Timeseries Based Candlesticks (OHLC)
plot_0 = vz.g_ohlc(p_ohlc=df_data)
# interactive plot with plotly (check visualizations.py)
# plot_0.show()
# --------------------------------------------------------------------------------------------- Notebook -- #
## Construct linear features (Autoregressive)
# -- Linear Features Engineering
lin_features = fn.linear_features(p_data=df_data, p_memory=7, p_target='co')
# data profile
lin_features_profile = fn.data_profile(p_data=lin_features.copy(), p_type='ts', p_mult=1)
# description
print(lin_features_profile)
# --------------------------------------------------------------------------------------------- Notebook -- #
## Scale linear features (robust)
# -- Linear Features Scaling
lin_features = fn.data_scaler(p_data=lin_features, p_trans='standard')
# data profile
lin_features_profile = fn.data_profile(p_data=lin_features.copy(), p_type='ts', p_mult=1)
# description
print(lin_features_profile)
# --------------------------------------------------------------------------------------------- Notebook -- #
## Use gplearn library
# --------------------------------------------------------------------------------------------- Notebook -- #
## Parameters for Symbolic Variable Generation Through Genetic Programming
# 'population': 15000, 'tournament': 2000, 'hof': 30, 'generations': 9, 'n_features': 20,
# paremeters for symbolic features generation process
symbolic_params = {'functions': ['sub', 'add', 'inv', 'mul', 'div', 'abs', 'log', 'sqrt'],
'population': 12000, 'tournament': 3000, 'hof': 30, 'generations': 5, 'n_features': 30,
'init_depth': (4, 10), 'init_method': 'half and half', 'parsimony': 0.001,
'constants': None,
'metric': 'pearson', 'metric_goal': 0.90,
'prob_cross': 0.4, 'prob_mutation_subtree': 0.5,
'prob_mutation_hoist': 0.05, 'prob_mutation_point': 0.05,
'max_samples': 1,
'verbose': True, 'parallelization': True, 'warm_start': True}
# --------------------------------------------------------------------------------------------- Notebook -- #
## Symbolic Features Engineering with Genetic Programming
# Target variable name
y_hat = 'co'
# Run process
genetic_prog = fn.genetic_programed_features(p_data=lin_features, p_target=y_hat, p_params=symbolic_params)
# Process description
sym_process = pd.DataFrame(genetic_prog['sym_data']['details'])
# --------------------------------------------------------------------------------------------- Notebook -- #
## Some special notes on this use case of gplearn
# fitness is a demean value transformation in gplearn, calculate .corr() to have original pearson value
# 'best_oob_fitness' == Out-of-bag error for best individual (not applicable for SymbolicTransformer)
# The sum of p_crossover, p_subtree_mutation, p_hoist_mutation and p_point_mutation should total to 1.0
# is possible to have repeated elements (symbolic features) because: low generations mostly
# to get some properties like raw_fitness_ and fitness_ which allow you to get the raw fitness and fitness
# (regularized by length) for each program
# max_samples not recomended for timeseries it does shuffle data
# warm_start = True for continuing evolution and not loose previous generation
# --------------------------------------------------------------------------------------------- Notebook -- #
## Get info of best programms
# best programs
best_progs = genetic_prog['sym_data']['best_programs']
best_progs
# --------------------------------------------------------------------------------------------- Notebook -- #
## Get info of best features
# symbolic features
sym_features = genetic_prog['sym_features']
# Feature description
# sym_features.describe()
# --------------------------------------------------------------------------------------------- Notebook -- #
## EXPERIMENT 1: Just Symbolic Features
exp_1 = sym_features.copy()
exp_1[y_hat] = lin_features[y_hat].copy()
exp_1 = exp_1.reindex(columns=sorted(list(exp_1.columns)))
# Data for Experiment 3
# exp_1.head()
# correlation matrix
exp_corr_p = exp_1.corr('pearson')
exp_corr_s = exp_1.corr('pearson')
# Plots generation
plot_1_p = vz.g_heat_corr(p_data=exp_corr_p, p_double=False)
plot_1_s = vz.g_heat_corr(p_data=exp_corr_s, p_double=False)
# show plot
# plot_1_p.show()
# --------------------------------------------------------------------------------------------- Notebook -- #
## EXPERIMENT 2: Original Data and Symbolic Features
exp_2 = pd.concat([lin_features.copy(), sym_features.copy()], axis=1)
# Data for Experiment 3
exp_2.head()
# correlation matrix
exp_corr = exp_2.corr('pearson')
exp_corr_t = exp_corr.where(np.tril(np.ones(exp_corr.shape)).astype(np.bool_))
corr_1 = vz.g_heat_corr(p_data=exp_corr, p_double=False)
# plot 2
plt.figure(figsize=(18, 18))
sns.heatmap(exp_corr_t, cmap='Blues', cbar=True, square=True, center=0.0,
annot=False, cbar_kws={'shrink':.95}, fmt='.2f')
# formatting
plt.rcParams["xtick.labelcolor"] = 'darkgrey'
plt.rcParams["ytick.labelcolor"] = 'darkgrey'
plt.rcParams["ytick.labelsize"] = 8
plt.rcParams["xtick.labelsize"] = 8
# show plot
# plt.show()
# --------------------------------------------------------------------------------------------- Notebook -- #
## EXPERIMENT 3: Just 'important' variables from Original Data & Symbolic Features
# get all the data (linear features and symbolic features)
exp_3 = exp_2.copy()
# Correlation with target variable most be >= condition_1
condition_1 = 0.10
# Absolute correlation among all variables most be <= condition_2
condition_2 = 0.5
# Correlation matrix
exp_3_corr = exp_3.corr('pearson')
no_ok_1 = list(exp_3.columns[abs(exp_3_corr[y_hat]) < condition_1])
exp_3_1 = exp_3.drop(no_ok_1, inplace=False, axis=1)
# Sub correlation matrix
exp_3_1_corr = exp_3_1.corr('spearman')
# Drop row and column name like target
exp_3_1_corr.drop(labels=y_hat, axis=0, inplace=True)
exp_3_1_corr.drop(labels=y_hat, axis=1, inplace=True)
# Transform to 1 all the elements below diagnoal and select the ones below the condition_2
upper_tri = exp_3_1_corr.where(np.triu(np.ones(exp_3_1_corr.shape), k=1).astype(bool))
no_ok_2 = [column for column in upper_tri.columns if any(abs(upper_tri[column]) > condition_2)]
exp_3_1_corr.drop(labels=no_ok_2, axis=0, inplace=True)
exp_3_1_corr.drop(labels=no_ok_2, axis=1, inplace=True)
exp_3_2_corr = exp_3_1_corr
# The most correlated to the target and the least correlated to each other
exp_3 = exp_3[['co'] + list(exp_3_2_corr.columns)]
# Data for Experiment 3
# exp_3.head()
# --------------------------------------------------------------------------------------------- Notebook -- #
### Correlation heatmaps
# layout grid
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(18,18))
# calculation for plot 1
exp_corr = exp_3.corr('pearson')
exp_corr_t = exp_corr.where(np.tril(np.ones(exp_corr.shape)).astype(np.bool_))
# plot 1
sns.heatmap(exp_corr_t, cmap='Blues', cbar=False, square=True, ax=ax1,
annot=True, annot_kws={'fontsize':8}, fmt='.2f')
ax1.set_title("|Pearson Correlation| > 0.10")
# calculation for plot 2
exp_corr = exp_3.corr('spearman')
exp_corr_t = exp_corr.where(np.tril(np.ones(exp_corr.shape)).astype(np.bool_))
# plot 2
sns.heatmap(exp_corr_t, cmap='Blues', cbar=False, square=True, ax=ax2,
annot=True, annot_kws={'fontsize':8}, fmt='.2f')
ax2.set_title("|spearman Correlation| < 0.50")
# formatting
plt.rcParams["xtick.labelcolor"] = 'darkgrey'
plt.rcParams["ytick.labelcolor"] = 'darkgrey'
plt.rcParams["ytick.labelsize"] = 8
plt.rcParams["xtick.labelsize"] = 8
# show plot
# plt.show()
# --------------------------------------------------------------------------------------------- Notebook -- #
## EXPERIMENT 4: top 10 'important' variables (original + symbolic) both using pearson and spearman
# get all the data (linear features and symbolic features)
exp_4 = exp_2.copy()
# Top N important variables (pearson and spearman)
n_important = 10
# use pearson coefficient
exp_4_p = exp_4.corr('pearson')
# use spearman coefficient
exp_4_s = exp_4.corr('spearman')
# create dataframes with top N features according to correlation criteria
exp_4_p = exp_4.copy()[abs(exp_4_p['co']).sort_values(ascending=False)[1:n_important].index]
exp_4_s = exp_4.copy()[abs(exp_4_s['co']).sort_values(ascending=False)[1:n_important].index]
exp_4 = pd.concat([exp_4 ['co'], exp_4_s, exp_4_p], axis=1)
# print top 5 rows
# exp_4.head()
# --------------------------------------------------------------------------------------------- Notebook -- #
### Correlation heatmaps
# layout grid
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(14,14))
# calculation for plot 1
exp_corr = exp_4.corr('pearson')
exp_corr_t = exp_corr.where(np.tril(np.ones(exp_corr.shape)).astype(np.bool_))
# plot 1
sns.heatmap(exp_corr_t, cmap='Blues', cbar=False, square=True, ax=ax1,
annot=True, annot_kws={'fontsize':8}, fmt='.2f')
ax1.set_title("Pearson Correlation")
# calculation for plot 2
exp_corr = exp_4.corr('spearman')
exp_corr_t = exp_corr.where(np.tril(np.ones(exp_corr.shape)).astype(np.bool_))
# plot 2
sns.heatmap(exp_corr_t, cmap='Blues', cbar=False, square=True, ax=ax2,
annot=True, annot_kws={'fontsize':8}, fmt='.2f')
ax2.set_title("spearman Correlation")
# formatting
plt.rcParams["xtick.labelcolor"] = 'darkgrey'
plt.rcParams["ytick.labelcolor"] = 'darkgrey'
plt.rcParams["ytick.labelsize"] = 8
plt.rcParams["xtick.labelsize"] = 8
# show plot
# plt.show()
# --------------------------------------------------------------------------------------------- Notebook -- #
## Models
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.optimizers import SGD
from sklearn.metrics import r2_score
experiments = {1: exp_1, 2: exp_2, 3: exp_3, 4: exp_4}
exp = 4
data = fn.data_split(p_data=experiments[exp], p_target='co', p_split=0.8)
x_train = data['train_x']
x_val = data['val_x']
y_train = data['train_y']
y_val = data['val_y']
learning_rate = 0.001
epochs = 500
batch = 16
neurons = x_train.shape[1]
# Neural network architecture
model = Sequential()
model.add(Dense(neurons, activation='sigmoid', input_dim=x_train.shape[1]))
model.add(Dense(neurons, activation='sigmoid'))
model.add(Dense(1, activation='linear'))
opt = SGD(lr=learning_rate)
model.compile(loss = 'mean_squared_error', optimizer=opt, metrics=['mse'])
# fit the model
model_history = model.fit(x_train, y_train, epochs=epochs, batch_size=batch,
validation_data=(x_val, y_val), verbose=2)
model_score_t = model.evaluate(x_train, y_train)
model_score_v = model.evaluate(x_val, y_val)
print('Train loss:', model_score_t[0])
print('Train mse:', model_score_t[1])
print('Val loss:', model_score_v[0])
print('Val mse:', model_score_v[1])
fig, ax = plt.subplots(1, 1, figsize=(10,6))
ax.plot(model_history.history['loss'], 'r', label='train')
ax.plot(model_history.history['val_loss'], 'b' ,label='val')
ax.set_xlabel(r'Epoch', fontsize=20)
ax.set_ylabel(r'Loss', fontsize=20)
ax.legend()
ax.tick_params(labelsize=20)
y_hat = model.predict(x_train)
R2_score = r2_score(y_train, y_hat)
y_hat_val = model.predict(x_val)
R2_score_val = r2_score(y_val, y_hat_val)
x_min, x_max = min(y_train),max(y_train)
x_line = np.linspace(x_min, x_max)
fig = plt.figure(figsize=(10,6))
plt.scatter(y_train,y_hat,label='Train Estimation')
plt.plot(x_line, x_line, 'k--', label='Perfect estimation')
plt.xlabel('Real output', fontsize=20)
plt.ylabel('Estimation output', fontsize=20)
plt.title('R^2=%0.4f'%R2_score, fontsize=20)
plt.legend()
plt.grid()
plt.show()
x_min, x_max = min(y_train),max(y_train)
x_line = np.linspace(x_min, x_max)
fig = plt.figure(figsize=(10,6))
plt.scatter(y_val, y_hat_val,label='Test Estimation')
plt.plot(x_line, x_line, 'k--', label='Perfect estimation')
plt.xlabel('Real output', fontsize=20)
plt.ylabel('Estimation output', fontsize=20)
plt.title('R^2=%0.4f'%R2_score_val, fontsize=20)
plt.legend()
plt.grid()
plt.show()