File size: 11,227 Bytes
2065fdc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""Options Pricing with ML - Neural network for option pricing/IV prediction."""
import numpy as np
import pandas as pd
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from typing import Dict, Tuple, Optional
import warnings
warnings.filterwarnings('ignore')


class OptionDataset(Dataset):
    """Dataset for option pricing"""
    def __init__(self, X: np.ndarray, y: np.ndarray):
        self.X = torch.FloatTensor(X)
        self.y = torch.FloatTensor(y).unsqueeze(1)
    
    def __len__(self):
        return len(self.X)
    
    def __getitem__(self, idx):
        return self.X[idx], self.y[idx]


class OptionPricingNN(nn.Module):
    """Neural network for option pricing"""
    def __init__(self, input_size: int, hidden_sizes: list = [256, 128, 64, 32]):
        super().__init__()
        
        layers = []
        prev_size = input_size
        for hidden_size in hidden_sizes:
            layers.extend([
                nn.Linear(prev_size, hidden_size),
                nn.ReLU(),
                nn.Dropout(0.2)
            ])
            prev_size = hidden_size
        
        layers.append(nn.Linear(prev_size, 1))
        self.network = nn.Sequential(*layers)
        
    def forward(self, x):
        return self.network(x)


class BlackScholes:
    """Analytical Black-Scholes for baseline comparison"""
    
    @staticmethod
    def d1(S, K, T, r, sigma):
        from scipy.stats import norm
        return (np.log(S / K) + (r + 0.5 * sigma ** 2) * T) / (sigma * np.sqrt(T))
    
    @staticmethod
    def d2(S, K, T, r, sigma):
        return BlackScholes.d1(S, K, T, r, sigma) - sigma * np.sqrt(T)
    
    @staticmethod
    def call_price(S, K, T, r, sigma):
        from scipy.stats import norm
        d1 = BlackScholes.d1(S, K, T, r, sigma)
        d2 = BlackScholes.d2(S, K, T, r, sigma)
        return S * norm.cdf(d1) - K * np.exp(-r * T) * norm.cdf(d2)
    
    @staticmethod
    def put_price(S, K, T, r, sigma):
        from scipy.stats import norm
        d1 = BlackScholes.d1(S, K, T, r, sigma)
        d2 = BlackScholes.d2(S, K, T, r, sigma)
        return K * np.exp(-r * T) * norm.cdf(-d2) - S * norm.cdf(-d1)
    
    @staticmethod
    def implied_volatility(price, S, K, T, r, option_type='call', tol=1e-5, max_iter=100):
        """Find implied volatility using Newton-Raphson"""
        sigma = 0.2  # Initial guess
        for _ in range(max_iter):
            if option_type == 'call':
                price_est = BlackScholes.call_price(S, K, T, r, sigma)
            else:
                price_est = BlackScholes.put_price(S, K, T, r, sigma)
            
            diff = price_est - price
            if abs(diff) < tol:
                return sigma
            
            # Vega
            from scipy.stats import norm
            d1 = BlackScholes.d1(S, K, T, r, sigma)
            vega = S * norm.pdf(d1) * np.sqrt(T)
            
            if vega < 1e-10:
                break
            
            sigma -= diff / vega
            sigma = max(sigma, 0.001)
        
        return sigma


class MLOptionsPricer:
    """ML-based options pricing engine"""
    
    def __init__(self, hidden_sizes: list = [256, 128, 64, 32],
                 device: str = 'cpu'):
        self.hidden_sizes = hidden_sizes
        self.device = torch.device(device)
        self.model = None
        self.bs = BlackScholes()
        
    def prepare_features(self, options_df: pd.DataFrame) -> np.ndarray:
        """
        Prepare features for ML model
        
        Expected columns: S, K, T, r, sigma_hist, option_type, 
                         S_lag_1, S_lag_2, ..., S_lag_20
        """
        features = []
        
        # Core features
        features.append(options_df['S'].values)
        features.append(options_df['K'].values)
        features.append(options_df['T'].values)
        features.append(options_df['r'].values)
        features.append(options_df['sigma_hist'].values)
        features.append((options_df['S'] / options_df['K']).values)  # Moneyness
        features.append(options_df['T'].values * 252)  # Days to expiry
        
        # Option type encoding
        features.append((options_df['option_type'] == 'call').astype(float).values)
        
        # Lag features (past 20 days of underlying price)
        for i in range(1, 21):
            col = f'S_lag_{i}'
            if col in options_df.columns:
                features.append(options_df[col].values)
        
        return np.column_stack(features)
    
    def fit(self, X_train: np.ndarray, y_train: np.ndarray,
            X_val: Optional[np.ndarray] = None, y_val: Optional[np.ndarray] = None,
            epochs: int = 100, batch_size: int = 256, lr: float = 1e-3) -> Dict:
        """Train the neural network"""
        input_size = X_train.shape[1]
        self.model = OptionPricingNN(input_size, self.hidden_sizes).to(self.device)
        
        train_dataset = OptionDataset(X_train, y_train)
        train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
        
        optimizer = torch.optim.Adam(self.model.parameters(), lr=lr)
        scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, patience=10)
        criterion = nn.MSELoss()
        
        metrics = {'train_loss': [], 'val_loss': [], 'val_mae': []}
        
        for epoch in range(epochs):
            self.model.train()
            epoch_loss = 0
            for X_batch, y_batch in train_loader:
                X_batch, y_batch = X_batch.to(self.device), y_batch.to(self.device)
                optimizer.zero_grad()
                pred = self.model(X_batch)
                loss = criterion(pred, y_batch)
                loss.backward()
                optimizer.step()
                epoch_loss += loss.item()
            
            avg_train_loss = epoch_loss / len(train_loader)
            metrics['train_loss'].append(avg_train_loss)
            
            # Validation
            if X_val is not None and y_val is not None:
                self.model.eval()
                with torch.no_grad():
                    X_val_t = torch.FloatTensor(X_val).to(self.device)
                    y_val_t = torch.FloatTensor(y_val).to(self.device)
                    val_pred = self.model(X_val_t)
                    val_loss = criterion(val_pred, y_val_t).item()
                    val_mae = torch.mean(torch.abs(val_pred - y_val_t)).item()
                    metrics['val_loss'].append(val_loss)
                    metrics['val_mae'].append(val_mae)
                
                scheduler.step(val_loss)
                
                if epoch % 10 == 0:
                    print(f"  Epoch {epoch}: train_loss={avg_train_loss:.6f}, "
                          f"val_loss={val_loss:.6f}, val_mae={val_mae:.4f}")
        
        return metrics
    
    def predict(self, X: np.ndarray) -> np.ndarray:
        """Predict option prices"""
        if self.model is None:
            raise ValueError("Model must be trained before prediction")
        
        self.model.eval()
        with torch.no_grad():
            X_t = torch.FloatTensor(X).to(self.device)
            pred = self.model(X_t).cpu().numpy().flatten()
        
        return pred
    
    def predict_iv(self, options_df: pd.DataFrame, market_prices: np.ndarray) -> np.ndarray:
        """
        Predict implied volatility by inverting the model
        Uses Black-Scholes as baseline and ML as correction
        """
        S = options_df['S'].values
        K = options_df['K'].values
        T = options_df['T'].values
        r = options_df['r'].values
        option_type = options_df['option_type'].values
        
        # Get ML prediction
        X = self.prepare_features(options_df)
        ml_price = self.predict(X)
        
        # Get Black-Scholes baseline
        bs_iv = np.array([
            self.bs.implied_volatility(
                market_prices[i], S[i], K[i], T[i], r[i], option_type[i]
            )
            for i in range(len(market_prices))
        ])
        
        # ML-adjusted IV: if ML price differs from market, adjust IV accordingly
        ml_iv = np.array([
            self.bs.implied_volatility(
                ml_price[i], S[i], K[i], T[i], r[i], option_type[i]
            )
            for i in range(len(ml_price))
        ])
        
        # Ensemble: weighted average
        ensemble_iv = 0.5 * bs_iv + 0.5 * ml_iv
        
        return ensemble_iv
    
    def detect_mispricing(self, options_df: pd.DataFrame, 
                          market_prices: np.ndarray,
                          threshold: float = 0.05) -> pd.DataFrame:
        """
        Detect mispriced options
        
        Returns options where |ML_price - market_price| / market_price > threshold
        """
        X = self.prepare_features(options_df)
        ml_prices = self.predict(X)
        
        mispricing = (ml_prices - market_prices) / market_prices
        
        result = options_df.copy()
        result['ml_price'] = ml_prices
        result['market_price'] = market_prices
        result['mispricing_pct'] = mispricing * 100
        result['signal'] = np.where(
            mispricing > threshold, 'OVERPRICED',
            np.where(mispricing < -threshold, 'UNDERPRICED', 'FAIR')
        )
        
        return result
    
    def generate_synthetic_options(self, n_samples: int = 10000,
                                    S_range: Tuple[float, float] = (50, 200),
                                    K_range: Tuple[float, float] = (50, 200),
                                    T_range: Tuple[float, float] = (0.01, 1.0),
                                    r_range: Tuple[float, float] = (0.01, 0.05),
                                    sigma_range: Tuple[float, float] = (0.1, 0.5)) -> pd.DataFrame:
        """Generate synthetic option data for training"""
        np.random.seed(42)
        
        S = np.random.uniform(*S_range, n_samples)
        K = np.random.uniform(*K_range, n_samples)
        T = np.random.uniform(*T_range, n_samples)
        r = np.random.uniform(*r_range, n_samples)
        sigma = np.random.uniform(*sigma_range, n_samples)
        option_type = np.random.choice(['call', 'put'], n_samples)
        
        # Generate lag features (simulated price history)
        lags = {}
        for i in range(1, 21):
            lags[f'S_lag_{i}'] = S * (1 + np.random.normal(0, 0.01, n_samples))
        
        # Calculate prices using Black-Scholes with noise
        prices = []
        for i in range(n_samples):
            if option_type[i] == 'call':
                price = self.bs.call_price(S[i], K[i], T[i], r[i], sigma[i])
            else:
                price = self.bs.put_price(S[i], K[i], T[i], r[i], sigma[i])
            # Add noise
            price *= (1 + np.random.normal(0, 0.02))
            prices.append(max(price, 0.01))
        
        df = pd.DataFrame({
            'S': S,
            'K': K,
            'T': T,
            'r': r,
            'sigma_hist': sigma,
            'option_type': option_type,
            'price': prices,
            **lags
        })
        
        return df