supervised_financial_model.py

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression # Example model
from sklearn.metrics import accuracy_score, classification_report

# (Placeholder for more sophisticated feature engineering)
def create_features(df: pd.DataFrame, target_lag: int = 1) -> pd.DataFrame:
    """
    Creates basic features for financial time series.
    - Lagged returns
    - Target variable (e.g., price goes up or down)
    """
    df_copy = df.copy()
    df_copy['returns'] = df_copy['Close'].pct_change()
    
    # Simple target: 1 if next day's close is higher, 0 otherwise
    df_copy['target'] = (df_copy['Close'].shift(-target_lag) > df_copy['Close']).astype(int)
    
    # Add more features: e.g., moving averages, RSI, MACD
    df_copy['ma5'] = df_copy['Close'].rolling(window=5).mean()
    df_copy['ma20'] = df_copy['Close'].rolling(window=20).mean()
    
    df_copy = df_copy.dropna()
    return df_copy

def preprocess_data_for_supervised(
    df: pd.DataFrame,
    features_list: list = ['returns', 'ma5', 'ma20'],
    target_col: str = 'target',
    test_size: float = 0.2,
    random_state: int = 42
):
    """
    Prepares data for supervised learning.
    - Selects features and target.
    - Splits data into training and testing sets.
    - Scales features.
    """
    X = df[features_list]
    y = df[target_col]
    
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size, random_state=random_state, shuffle=False) # Time series data, so no shuffle
    
    scaler = StandardScaler()
    X_train_scaled = scaler.fit_transform(X_train)
    X_test_scaled = scaler.transform(X_test)
    
    return X_train_scaled, X_test_scaled, y_train, y_test, scaler

def train_supervised_model(X_train, y_train, model_type='logistic_regression', model_params=None):
    """
    Trains a supervised learning model.
    """
    if model_params is None:
        model_params = {}
        
    if model_type == 'logistic_regression':
        model = LogisticRegression(**model_params, random_state=42, max_iter=1000) # Added max_iter
    # Add other model types here (e.g., SVM, RandomForest, GradientBoosting)
    # elif model_type == 'random_forest':
    #     from sklearn.ensemble import RandomForestClassifier
    #     model = RandomForestClassifier(**model_params, random_state=42)
    else:
        raise ValueError(f"Unsupported model type: {model_type}")
        
    model.fit(X_train, y_train)
    return model

def evaluate_supervised_model(model, X_test, y_test):
    """
    Evaluates the trained supervised model.
    """
    predictions = model.predict(X_test)
    accuracy = accuracy_score(y_test, predictions)
    report = classification_report(y_test, predictions)
    
    print(f"Model Accuracy: {accuracy:.4f}")
    print("Classification Report:")
    print(report)
    
    return accuracy, report, predictions

if __name__ == '__main__':
    # This is an example of how to use the functions.
    # You'll need to integrate this with your data fetching agent.
    
    # 1. Fetch data (using the financial_data_agent you modified)
    # from agents.financial_data_agent import fetch_historical_ohlcv
    # raw_data = fetch_historical_ohlcv("AAPL", period="1y", interval="1d")
    
    # For demonstration, creating a dummy DataFrame:
    dates = pd.date_range(start='2023-01-01', periods=200, freq='B')
    data = np.random.rand(200, 5) * 100 + 100
    raw_data = pd.DataFrame(data, index=dates, columns=['Open', 'High', 'Low', 'Close', 'Volume'])
    raw_data['Close'] = raw_data['Close'] + np.sin(np.linspace(0, 10, 200)) * 10 # Add some trend

    if not raw_data.empty:
        # 2. Create features
        featured_data = create_features(raw_data)
        
        if not featured_data.empty:
            # 3. Preprocess data
            X_train_scaled, X_test_scaled, y_train, y_test, scaler = preprocess_data_for_supervised(
                featured_data, 
                features_list=['returns', 'ma5', 'ma20'] # Ensure these features exist
            )
            
            # 4. Train model
            print("Training supervised model...")
            trained_model = train_supervised_model(X_train_scaled, y_train)
            print("Model trained.")
            
            # 5. Evaluate model
            print("\nEvaluating model...")
            evaluate_supervised_model(trained_model, X_test_scaled, y_test)
        else:
            print("Featured data is empty. Check feature creation.")
    else:
        print("Raw data is empty. Check data fetching.")