train.py

import tensorflow as tf
from tensorflow.keras import layers
import pandas as pd
import numpy as np
from typing import Tuple, List
import logging
from datetime import datetime
from pathlib import Path
import json
from sklearn.preprocessing import MinMaxScaler
from ta.trend import SMAIndicator, EMAIndicator, MACD
from ta.momentum import RSIIndicator
from ta.volatility import BollingerBands

# Set up logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)

class DataPreprocessor:
    """Handles data loading and preprocessing"""
    
    def __init__(self, config_path: str = 'training_config/hyperparameters.json'):
        with open(config_path) as f:
            self.config = json.load(f)
        self.scalers = {}
        
    def load_data(self, data_path: str) -> pd.DataFrame:
        """Load data from CSV and add technical indicators"""
        df = pd.read_csv(data_path)
        df['timestamp'] = pd.to_datetime(df['timestamp'])
        df = df.sort_values('timestamp')
        
        # Add technical indicators
        df = self.add_technical_indicators(df)
        
        return df
        
    def add_technical_indicators(self, df: pd.DataFrame) -> pd.DataFrame:
        """Add technical analysis indicators"""
        # SMA
        df['sma_20'] = SMAIndicator(close=df['price'], window=20).sma_indicator()
        df['sma_50'] = SMAIndicator(close=df['price'], window=50).sma_indicator()
        
        # EMA
        df['ema_20'] = EMAIndicator(close=df['price'], window=20).ema_indicator()
        
        # MACD
        macd = MACD(close=df['price'])
        df['macd'] = macd.macd()
        df['macd_signal'] = macd.macd_signal()
        
        # RSI
        df['rsi'] = RSIIndicator(close=df['price']).rsi()
        
        # Bollinger Bands
        bb = BollingerBands(close=df['price'])
        df['bb_high'] = bb.bollinger_hband()
        df['bb_low'] = bb.bollinger_lband()
        
        return df
        
    def prepare_sequences(self, df: pd.DataFrame) -> Tuple[np.ndarray, np.ndarray]:
        """Create sequences for training"""
        sequence_length = self.config['data']['sequence_length']
        
        # Scale features
        for column in df.select_dtypes(include=[np.number]).columns:
            if column not in self.scalers:
                self.scalers[column] = MinMaxScaler()
            df[column] = self.scalers[column].fit_transform(df[[column]])
        
        # Create sequences
        sequences = []
        targets = []
        
        for i in range(len(df) - sequence_length):
            sequence = df.iloc[i:i + sequence_length]
            target = df.iloc[i + sequence_length]['price']
            sequences.append(sequence)
            targets.append(target)
            
        return np.array(sequences), np.array(targets)

class TransformerBlock(layers.Layer):
    """Transformer block with multi-head attention"""
    
    def __init__(self, embed_dim, num_heads, ff_dim, rate=0.1):
        super().__init__()
        self.att = layers.MultiHeadAttention(num_heads=num_heads, key_dim=embed_dim)
        self.ffn = tf.keras.Sequential([
            layers.Dense(ff_dim, activation="relu"),
            layers.Dense(embed_dim),
        ])
        self.layernorm1 = layers.LayerNormalization(epsilon=1e-6)
        self.layernorm2 = layers.LayerNormalization(epsilon=1e-6)
        self.dropout1 = layers.Dropout(rate)
        self.dropout2 = layers.Dropout(rate)

    def call(self, inputs, training):
        attn_output = self.att(inputs, inputs)
        attn_output = self.dropout1(attn_output, training=training)
        out1 = self.layernorm1(inputs + attn_output)
        ffn_output = self.ffn(out1)
        ffn_output = self.dropout2(ffn_output, training=training)
        return self.layernorm2(out1 + ffn_output)

class CryptoTransformer(tf.keras.Model):
    """Main transformer model for cryptocurrency prediction"""
    
    def __init__(self, config_path: str = 'training_config/hyperparameters.json'):
        super().__init__()
        
        with open(config_path) as f:
            self.config = json.load(f)
        
        # Model parameters
        self.num_layers = self.config['model']['n_layers']
        self.d_model = self.config['model']['d_model']
        self.num_heads = self.config['model']['n_heads']
        self.ff_dim = self.config['model']['d_ff']
        self.dropout = self.config['model']['dropout']
        
        # Layers
        self.transformer_blocks = [
            TransformerBlock(self.d_model, self.num_heads, self.ff_dim, self.dropout)
            for _ in range(self.num_layers)
        ]
        self.dropout = layers.Dropout(self.dropout)
        self.dense = layers.Dense(1)  # Final prediction layer
        
    def call(self, inputs):
        x = inputs
        for transformer_block in self.transformer_blocks:
            x = transformer_block(x)
        x = layers.GlobalAveragePooling1D()(x)
        x = self.dropout(x)
        return self.dense(x)

def train_model():
    """Main training function"""
    logger.info("Starting model training")
    
    # Initialize preprocessor and load data
    preprocessor = DataPreprocessor()
    df = preprocessor.load_data('data/training/kraken_trades.csv')
    
    # Prepare sequences
    X, y = preprocessor.prepare_sequences(df)
    
    # Split data
    train_size = int(0.8 * len(X))
    X_train, X_test = X[:train_size], X[train_size:]
    y_train, y_test = y[:train_size], y[train_size:]
    
    # Initialize model
    model = CryptoTransformer()
    
    # Compile model
    optimizer = tf.keras.optimizers.Adam(learning_rate=1e-4)
    model.compile(
        optimizer=optimizer,
        loss='mse',
        metrics=['mae']
    )
    
    # Train model
    history = model.fit(
        X_train, y_train,
        epochs=100,
        batch_size=32,
        validation_data=(X_test, y_test),
        callbacks=[
            tf.keras.callbacks.EarlyStopping(
                monitor='val_loss',
                patience=10,
                restore_best_weights=True
            ),
            tf.keras.callbacks.ModelCheckpoint(
                'models/crypto_transformer_{epoch}.h5',
                save_best_only=True,
                monitor='val_loss'
            ),
            tf.keras.callbacks.TensorBoard(log_dir='logs')
        ]
    )
    
    # Save final model
    model.save('models/crypto_transformer_final')
    
    # Save training history
    pd.DataFrame(history.history).to_csv('models/training_history.csv')
    
    logger.info("Training completed")
    return model, history

if __name__ == "__main__":
    # Create necessary directories
    Path('models').mkdir(exist_ok=True)
    Path('logs').mkdir(exist_ok=True)
    
    # Train model
    model, history = train_model()