-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
252 lines (197 loc) · 8.83 KB
/
main.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
import pandas as pd
import numpy as np
import xgboost as xgb
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, precision_score, recall_score, roc_auc_score
from sklearn.preprocessing import StandardScaler, LabelEncoder
import matplotlib.pyplot as plt
import joblib
import ta
from skopt import BayesSearchCV
import logging
# Set up logging
logging.basicConfig(level=logging.INFO,
format='%(asctime)s - %(levelname)s - %(message)s')
# Load the data
def load_data(filename):
try:
data = pd.read_csv(filename)
logging.info(f'Successfully loaded data from {filename}')
except Exception as e:
logging.error(f'Error loading data from {filename}: {e}')
data = None
return data
# Load the data Training
def load_dataT(filenames):
dataframes = []
for filename in filenames:
try:
data = pd.read_csv(filename)
dataframes.append(data)
logging.info(f'Successfully loaded data from {filename}')
except Exception as e:
logging.error(f'Error loading data from {filename}: {e}')
return pd.concat(dataframes, ignore_index=True)
# Preprocess the data
def preprocess_data(data):
data = data.dropna()
le = LabelEncoder()
for column in data.columns:
if data[column].dtype == 'object':
data[column] = le.fit_transform(data[column])
scaler = StandardScaler()
data = pd.DataFrame(scaler.fit_transform(data), columns=data.columns)
return data
def generate_features_and_labels(data):
# Ensure that the index is a datetime
data.index = pd.to_datetime(data.index)
# Calculate EMA
data['EMA_5'] = ta.trend.ema_indicator(data['Close'], window=5)
# Calculate lagged features
data['Close_lag1'] = data['Close'].shift(1)
data['Volume_lag1'] = data['Volume'].shift(1)
# Calculate MACD
macd_indicator = ta.trend.MACD(data['Close'], window_fast=3, window_slow=9)
data['MACD_line'] = macd_indicator.macd()
data['Signal_line'] = macd_indicator.macd_signal()
# Calculate RSI
data['RSI_6'] = ta.momentum.RSIIndicator(data['Close'], window=6).rsi()
# Calculate Bollinger Bands
bollinger = ta.volatility.BollingerBands(data['Close'], window=5, window_dev=2)
data['Upper_BB'] = bollinger.bollinger_hband()
data['Lower_BB'] = bollinger.bollinger_lband()
# Calculate VWAP (Volume Weighted Average Price)
data['Volume'] = data['Volume'].astype(float) # Ensure Volume is float
data['VWAP'] = (data['Volume'] * data['Close']).cumsum() / data['Volume'].cumsum()
# Calculate ATR and Momentum
data['ATR_14'] = ta.volatility.average_true_range(data['High'], data['Low'], data['Close'], window=14)
data['Momentum_3'] = ta.momentum.roc(data['Close'], window=3)
# Simple head-and-shoulders pattern
data['Head-Shoulders'] = 0
for i in range(2, len(data) - 2):
if (data['Close'].iloc[i-2] < data['Close'].iloc[i-1]) and (data['Close'].iloc[i-1] > data['Close'].iloc[i]) and (data['Close'].iloc[i] < data['Close'].iloc[i+1]) and (data['Close'].iloc[i+1] < data['Close'].iloc[i+2]):
data['Head-Shoulders'].iloc[i] = 1
# Define label (1 when MACD crosses above signal line and Close crosses above EMA, 0 otherwise)
data['Label'] = 0
data.loc[(data['MACD_line'] > data['Signal_line']) & (data['Close'] > data['EMA_5']), 'Label'] = 1
data.loc[(data['MACD_line'] < data['Signal_line']) & (data['Close'] < data['EMA_5']), 'Label'] = 0
# Drop rows with NaN values
data = data.dropna()
# Split data into features and labels
features = data[['Close', 'EMA_5', 'MACD_line', 'Signal_line', 'RSI_6', 'Upper_BB', 'Lower_BB', 'VWAP', 'Head-Shoulders', 'Close_lag1', 'Volume_lag1', 'ATR_14', 'Momentum_3']]
labels = data['Label']
return features, labels
def train_model(features, labels):
X_train, X_test, y_train, y_test = train_test_split(
features, labels, test_size=0.2, random_state=42, stratify=labels)
# Define the sample weights
sample_weights = np.zeros(len(y_train))
sample_weights[y_train == 1] = 1 # Give weight of 1 to the positive class
sample_weights[y_train == 0] = 0.5 # Give smaller weight to the negative class
params = {
'max_depth': (3, 10),
'gamma': (0, 5),
'colsample_bytree': (0.5, 1),
'learning_rate': (0.01, 0.05),
'n_estimators': (500, 1000),
'min_child_weight': (1, 10),
'subsample': (0.5, 1),
'reg_lambda': (0.01, 1), # Added L2 regularization
'reg_alpha': (0.01, 1), # Added L1 regularization
'tree_method': ['gpu_hist'],
'n_jobs': [-1]
}
model = xgb.XGBClassifier()
bayes_search = BayesSearchCV(model, search_spaces=params,
n_iter=150, scoring='roc_auc', n_jobs=-1, cv=7, verbose=1, random_state=42)
bayes_search.fit(X_train, y_train, sample_weight=sample_weights)
model = bayes_search.best_estimator_
joblib.dump(model, '5momo101.pkl')
# Save the model's parameters
with open('model_params.txt', 'w') as file:
file.write(str(bayes_search.best_params_))
# Generate signals and evaluate the strategy
signals = generate_signals(model, X_test)
accuracy, precision, recall, roc_auc = evaluate_strategy(signals, y_test)
# Save the model's performance metrics
with open('model_metrics.txt', 'w') as file:
file.write(
f'Accuracy: {accuracy}, Precision: {precision}, Recall: {recall}, AUC-ROC: {roc_auc}')
return model, X_test, y_test
# Generate signals
def generate_signals(model, X_test):
predictions = model.predict(X_test)
signals = ['Buy' if prediction == 1 else 'Sell' for prediction in predictions]
# Ensure that the strategy is "Buy-Hold-Sell"
last_action = None
for i in range(len(signals)):
if last_action == signals[i]:
signals[i] = 'Hold'
else:
last_action = signals[i]
return signals
# Evaluate the strategy
def evaluate_strategy(signals, y_test):
labels = [1 if signal == 'Buy' else 0 for signal in signals]
accuracy = accuracy_score(y_test, labels)
precision = precision_score(y_test, labels, zero_division=1)
recall = recall_score(y_test, labels)
roc_auc = roc_auc_score(y_test, labels)
return accuracy, precision, recall, roc_auc
# Visualize the results
def visualize_results(signals, data):
plt.plot(data.index, data['Close'])
buy_signals = [i for i, signal in enumerate(signals) if signal == 'Buy']
plt.scatter(data.index[buy_signals], data['Close']
[buy_signals], color='green')
sell_signals = [i for i, signal in enumerate(signals) if signal == 'Sell']
plt.scatter(data.index[sell_signals], data['Close']
[sell_signals], color='red')
plt.show()
# Main function for generating signals
def main_generate_signals():
# Load and preprocess data
data = load_data('./data/BTCUSDT_5m.csv')
data = preprocess_data(data)
features, labels = generate_features_and_labels(data)
# Load model
model = load_model('./models/5momo101.pkl')
window_size = 20 # Set the size of the rolling window
signals = []
# Generate signals
for i in range(window_size, len(features)):
X_test = features.iloc[i-window_size:i]
signal = generate_signals(model, X_test)
signals.append(signal[-1]) # Store the most recent signal
# Transform list into a Pandas Series
signals = pd.Series(signals, index=features.index[window_size:])
# Ensure the first action is a 'Buy'
first_buy_index = signals[signals == 'Buy'].first_valid_index()
signals = signals.loc[first_buy_index:]
# Trim the labels to match the signals length
labels = labels.loc[signals.index]
# Evaluate strategy
accuracy, precision, recall, roc_auc = evaluate_strategy(signals, labels)
print(
f'Accuracy: {accuracy}, Precision: {precision}, Recall: {recall}, AUC-ROC: {roc_auc}')
# Visualize results
visualize_results(signals, data)
# Load the model
def load_model(model_path):
model = joblib.load(model_path)
return model
# # Main function for training
def main_train():
data = load_dataT(['./data/BTCUSDT_3m.csv', './data/ETHUSDT_3m.csv', './data/BNBUSDT_3m.csv', './data/ADAUSDT_3m.csv', './data/LTCUSDT_3m.csv', './data/XRPUSDT_3m.csv']) # Add more paths as needed
data = preprocess_data(data)
features, labels = generate_features_and_labels(data)
model, X_test, y_test = train_model(features, labels)
print('Model trained and saved successfully.')
if __name__ == '__main__':
choice = input("Enter '1' to train the model or '2' to generate signals: ")
if choice.lower() == '1':
main_train()
elif choice.lower() == '2':
main_generate_signals()
else:
print("Invalid choice. Please enter either '1' or '2'.")