Spaces:
Build error
Build error
File size: 9,343 Bytes
f828b45 |
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 |
import os
import pickle
import joblib
import pandas as pd
import numpy as np
from statsmodels.tsa.arima.model import ARIMA
from sklearn.preprocessing import MinMaxScaler
import warnings
warnings.filterwarnings('ignore')
class StockForecaster:
"""
A comprehensive stock forecasting class that combines ARIMA and LSTM models.
"""
def __init__(self):
self.arima_model = None
self.lstm_model = None
self.scaler = MinMaxScaler()
self.optimal_arima_params = (1, 1, 1) # Default parameters
def preprocess_data(self, data):
"""
Preprocess stock data for forecasting.
Args:
data (pd.DataFrame): Stock data with 'date' and 'close' columns
Returns:
pd.Series: Processed time series data
"""
if isinstance(data, pd.DataFrame):
if 'date' in data.columns:
data['date'] = pd.to_datetime(data['date'])
data = data.set_index('date').sort_index()
if 'close' in data.columns:
return data['close'].dropna()
return data.dropna()
def find_optimal_arima_params(self, ts_data, max_p=3, max_d=2, max_q=3):
"""
Find optimal ARIMA parameters using AIC criterion.
Args:
ts_data (pd.Series): Time series data
max_p, max_d, max_q (int): Maximum values for ARIMA parameters
Returns:
tuple: Optimal (p, d, q) parameters
"""
best_aic = np.inf
best_params = (1, 1, 1)
for p in range(max_p + 1):
for d in range(max_d + 1):
for q in range(max_q + 1):
try:
model = ARIMA(ts_data, order=(p, d, q))
fitted_model = model.fit()
aic = fitted_model.aic
if aic < best_aic:
best_aic = aic
best_params = (p, d, q)
except:
continue
self.optimal_arima_params = best_params
return best_params
def train_arima(self, ts_data):
"""
Train ARIMA model on time series data.
Args:
ts_data (pd.Series): Time series data
Returns:
statsmodels.ARIMAResults: Fitted ARIMA model
"""
try:
# Find optimal parameters if not set
if self.optimal_arima_params == (1, 1, 1):
self.find_optimal_arima_params(ts_data)
# Fit ARIMA model
arima_model = ARIMA(ts_data, order=self.optimal_arima_params)
self.arima_model = arima_model.fit()
return self.arima_model
except Exception as e:
print(f"ARIMA training error: {e}")
return None
def create_lstm_sequences(self, data, sequence_length=60):
"""
Create sequences for LSTM training.
Args:
data (np.array): Scaled time series data
sequence_length (int): Length of input sequences
Returns:
tuple: (X, y) arrays for LSTM training
"""
X, y = [], []
for i in range(sequence_length, len(data)):
X.append(data[i-sequence_length:i])
y.append(data[i])
return np.array(X), np.array(y)
def train_simple_lstm(self, ts_data, sequence_length=60):
"""
Train a simplified LSTM model or use trend-based prediction.
Args:
ts_data (pd.Series): Time series data
sequence_length (int): Sequence length for LSTM
Returns:
dict: Model information and scaler
"""
try:
# Scale the data
scaled_data = self.scaler.fit_transform(ts_data.values.reshape(-1, 1))
# For demo purposes, we'll use a trend-based approach
# In production, you'd train an actual LSTM here
self.lstm_model = {
'type': 'trend_based',
'data': scaled_data,
'sequence_length': sequence_length
}
return self.lstm_model
except Exception as e:
print(f"LSTM training error: {e}")
return None
def forecast_arima(self, steps=30):
"""
Generate ARIMA forecast.
Args:
steps (int): Number of steps to forecast
Returns:
np.array: Forecasted values
"""
if self.arima_model is None:
raise ValueError("ARIMA model not trained. Call train_arima first.")
try:
forecast = self.arima_model.forecast(steps=steps)
return forecast.values if hasattr(forecast, 'values') else forecast
except Exception as e:
print(f"ARIMA forecast error: {e}")
return None
def forecast_lstm(self, steps=30):
"""
Generate LSTM forecast (simplified trend-based approach).
Args:
steps (int): Number of steps to forecast
Returns:
np.array: Forecasted values
"""
if self.lstm_model is None:
raise ValueError("LSTM model not trained. Call train_simple_lstm first.")
try:
# Simple trend-based forecast for demo
scaled_data = self.lstm_model['data']
sequence_length = self.lstm_model['sequence_length']
# Calculate recent trend
recent_data = scaled_data[-min(10, len(scaled_data)):]
recent_trend = np.mean(np.diff(recent_data.flatten()))
# Generate forecast
last_value = scaled_data[-1][0]
forecast_scaled = []
for i in range(steps):
next_val = last_value + recent_trend * (i + 1) * 0.1
forecast_scaled.append([next_val])
# Inverse transform
forecast = self.scaler.inverse_transform(forecast_scaled).flatten()
return forecast
except Exception as e:
print(f"LSTM forecast error: {e}")
return None
def save_models(self, filepath_prefix):
"""
Save trained models to disk.
Args:
filepath_prefix (str): Prefix for saved model files
"""
try:
if self.arima_model is not None:
with open(f"{filepath_prefix}_arima.pkl", 'wb') as f:
pickle.dump(self.arima_model, f)
if self.lstm_model is not None:
joblib.dump(self.lstm_model, f"{filepath_prefix}_lstm.pkl")
joblib.dump(self.scaler, f"{filepath_prefix}_scaler.pkl")
print(f"Models saved with prefix: {filepath_prefix}")
except Exception as e:
print(f"Error saving models: {e}")
def load_models(self, filepath_prefix):
"""
Load trained models from disk.
Args:
filepath_prefix (str): Prefix for saved model files
"""
try:
# Load ARIMA model
arima_path = f"{filepath_prefix}_arima.pkl"
if os.path.exists(arima_path):
with open(arima_path, 'rb') as f:
self.arima_model = pickle.load(f)
# Load LSTM model
lstm_path = f"{filepath_prefix}_lstm.pkl"
if os.path.exists(lstm_path):
self.lstm_model = joblib.load(lstm_path)
# Load scaler
scaler_path = f"{filepath_prefix}_scaler.pkl"
if os.path.exists(scaler_path):
self.scaler = joblib.load(scaler_path)
print(f"Models loaded from prefix: {filepath_prefix}")
except Exception as e:
print(f"Error loading models: {e}")
# Example usage and model training script
if __name__ == "__main__":
# This would be run to pre-train models
forecaster = StockForecaster()
# Example with sample data
dates = pd.date_range('2020-01-01', periods=1000, freq='D')
prices = 100 + np.cumsum(np.random.randn(1000) * 0.5)
sample_data = pd.DataFrame({
'date': dates,
'close': prices
})
# Preprocess and train
ts_data = forecaster.preprocess_data(sample_data)
print("Training ARIMA model...")
forecaster.train_arima(ts_data)
print("Training LSTM model...")
forecaster.train_simple_lstm(ts_data)
# Generate forecasts
arima_forecast = forecaster.forecast_arima(30)
lstm_forecast = forecaster.forecast_lstm(30)
print(f"ARIMA forecast shape: {arima_forecast.shape if arima_forecast is not None else 'None'}")
print(f"LSTM forecast shape: {lstm_forecast.shape if lstm_forecast is not None else 'None'}")
# Save models
forecaster.save_models("models/stock_forecaster") |