#!/usr/bin/env python3
"""
Financial AI Predictor
======================

Sistema di predizione finanziaria basato su:
- Modello generalista pre-addestrato (scikit-learn)
- Layer finanziario specializzato
- Dati real-time da Yahoo Finance
- Feature engineering avanzato
- Validazione robusta

Author: Financial AI Research
License: Educational Use Only
"""

import warnings
import numpy as np
import pandas as pd
import yfinance as yf
import gradio as gr
import plotly.graph_objects as go
import plotly.express as px
from datetime import datetime, timedelta
import ta
from sklearn.ensemble import RandomForestRegressor, GradientBoostingRegressor
from sklearn.linear_model import Ridge
from sklearn.preprocessing import StandardScaler, RobustScaler
from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
from sklearn.model_selection import TimeSeriesSplit
import joblib
import json
from typing import Dict, List, Tuple, Optional

warnings.filterwarnings('ignore')

class FinancialFeatureExtractor:
    """Estrae feature finanziarie avanzate dai dati di mercato"""
    
    def __init__(self):
        self.feature_names = []
    
    def extract_technical_features(self, data: pd.DataFrame) -> pd.DataFrame:
        """Estrae indicatori tecnici avanzati"""
        features = pd.DataFrame(index=data.index)
        
        # Price-based features
        features['returns'] = data['Close'].pct_change()
        features['log_returns'] = np.log(data['Close'] / data['Close'].shift(1))
        features['price_momentum_5'] = data['Close'] / data['Close'].shift(5) - 1
        features['price_momentum_20'] = data['Close'] / data['Close'].shift(20) - 1
        
        # Moving averages and ratios
        for period in [5, 10, 20, 50]:
            ma = data['Close'].rolling(period).mean()
            features[f'ma_{period}_ratio'] = data['Close'] / ma
            features[f'ma_{period}_slope'] = ma.pct_change(5)
        
        # Volatility features
        features['volatility_5'] = features['returns'].rolling(5).std()
        features['volatility_20'] = features['returns'].rolling(20).std()
        features['volatility_ratio'] = features['volatility_5'] / features['volatility_20']
        
        # Volume features
        features['volume_sma'] = data['Volume'].rolling(20).mean()
        features['volume_ratio'] = data['Volume'] / features['volume_sma']
        features['volume_momentum'] = data['Volume'].pct_change(5)
        
        # Price-Volume features
        features['price_volume_trend'] = (data['Close'].pct_change() * 
                                        np.log(data['Volume'] + 1))
        
        return features
    
    def extract_ta_features(self, data: pd.DataFrame) -> pd.DataFrame:
        """Estrae indicatori tecnici usando la libreria TA"""
        features = pd.DataFrame(index=data.index)
        
        try:
            # Trend indicators
            features['sma_20'] = ta.trend.sma_indicator(data['Close'], window=20)
            features['ema_12'] = ta.trend.ema_indicator(data['Close'], window=12)
            features['ema_26'] = ta.trend.ema_indicator(data['Close'], window=26)
            features['macd'] = ta.trend.macd_diff(data['Close'])
            features['adx'] = ta.trend.adx(data['High'], data['Low'], data['Close'])
            
            # Momentum indicators
            features['rsi'] = ta.momentum.rsi(data['Close'])
            features['stoch'] = ta.momentum.stoch(data['High'], data['Low'], data['Close'])
            features['williams_r'] = ta.momentum.williams_r(data['High'], data['Low'], data['Close'])
            
            # Volatility indicators
            bb = ta.volatility.BollingerBands(data['Close'])
            features['bb_high'] = bb.bollinger_hband()
            features['bb_low'] = bb.bollinger_lband()
            features['bb_width'] = (features['bb_high'] - features['bb_low']) / data['Close']
            features['bb_position'] = (data['Close'] - features['bb_low']) / (features['bb_high'] - features['bb_low'])
            
            # Volume indicators
            features['obv'] = ta.volume.on_balance_volume(data['Close'], data['Volume'])
            features['cmf'] = ta.volume.chaikin_money_flow(data['High'], data['Low'], data['Close'], data['Volume'])
            features['vwap'] = ta.volume.volume_weighted_average_price(data['High'], data['Low'], data['Close'], data['Volume'])
            
        except Exception as e:
            print(f"Warning: Some TA features failed: {e}")
        
        return features
    
    def extract_market_regime_features(self, data: pd.DataFrame) -> pd.DataFrame:
        """Estrae feature di regime di mercato"""
        features = pd.DataFrame(index=data.index)
        
        # Trend strength
        returns = data['Close'].pct_change()
        features['trend_strength'] = returns.rolling(20).mean() / returns.rolling(20).std()
        
        # Market state indicators
        features['high_low_ratio'] = (data['High'] - data['Low']) / data['Close']
        features['close_position'] = (data['Close'] - data['Low']) / (data['High'] - data['Low'])
        
        # Volatility regime
        vol_20 = returns.rolling(20).std()
        vol_60 = returns.rolling(60).std()
        features['vol_regime'] = vol_20 / vol_60
        
        # Gap features
        features['gap'] = (data['Open'] - data['Close'].shift(1)) / data['Close'].shift(1)
        features['gap_filled'] = np.where(
            features['gap'] > 0,
            (data['Low'] <= data['Close'].shift(1)).astype(int),
            (data['High'] >= data['Close'].shift(1)).astype(int)
        )
        
        return features
    
    def extract_all_features(self, data: pd.DataFrame) -> pd.DataFrame:
        """Estrae tutte le feature"""
        print("📊 Extracting technical features...")
        tech_features = self.extract_technical_features(data)
        
        print("📈 Extracting TA indicators...")
        ta_features = self.extract_ta_features(data)
        
        print("🌊 Extracting market regime features...")
        regime_features = self.extract_market_regime_features(data)
        
        # Combina tutte le feature
        all_features = pd.concat([tech_features, ta_features, regime_features], axis=1)
        
        # Rimuovi feature con troppi NaN
        all_features = all_features.loc[:, all_features.isnull().mean() < 0.5]
        
        # Forward fill e backward fill
        all_features = all_features.fillna(method='ffill').fillna(method='bfill')
        
        # Rimuovi outliers estremi
        for col in all_features.columns:
            if all_features[col].dtype in ['float64', 'int64']:
                q99 = all_features[col].quantile(0.99)
                q01 = all_features[col].quantile(0.01)
                all_features[col] = all_features[col].clip(lower=q01, upper=q99)
        
        self.feature_names = all_features.columns.tolist()
        print(f"✅ Extracted {len(self.feature_names)} features")
        
        return all_features

class FinancialPredictor:
    """Sistema di predizione finanziaria con modelli generalisti"""
    
    def __init__(self):
        self.models = {}
        self.scalers = {}
        self.feature_extractor = FinancialFeatureExtractor()
        self.is_trained = False
        self.feature_importance = {}
        self.validation_scores = {}
        
    def create_models(self):
        """Crea i modelli generalisti"""
        self.models = {
            'random_forest': RandomForestRegressor(
                n_estimators=200,
                max_depth=15,
                min_samples_split=5,
                min_samples_leaf=2,
                random_state=42,
                n_jobs=-1
            ),
            'gradient_boost': GradientBoostingRegressor(
                n_estimators=150,
                max_depth=8,
                learning_rate=0.1,
                subsample=0.8,
                random_state=42
            ),
            'ridge': Ridge(
                alpha=1.0,
                random_state=42
            )
        }
        
        # Scaler robusto per gestire outliers
        self.scalers = {
            'robust': RobustScaler(),
            'standard': StandardScaler()
        }
    
    def prepare_data(self, symbol: str, period: str = "2y") -> Tuple[pd.DataFrame, pd.DataFrame]:
        """Prepara i dati per il training"""
        print(f"📥 Fetching data for {symbol}...")
        
        # Fetch data
        ticker = yf.Ticker(symbol)
        data = ticker.history(period=period)
        
        if data is None or len(data) < 100:
            raise ValueError(f"Insufficient data for {symbol}")
        
        print(f"📊 Data shape: {data.shape}")
        
        # Extract features
        features = self.feature_extractor.extract_all_features(data)
        
        # Create targets (predicting next day return)
        targets = pd.DataFrame(index=data.index)
        targets['next_return'] = data['Close'].pct_change().shift(-1)
        targets['next_direction'] = (targets['next_return'] > 0).astype(int)
        targets['next_volatility'] = targets['next_return'].rolling(5).std().shift(-1)
        
        # Align data
        common_index = features.index.intersection(targets.index)
        features = features.loc[common_index]
        targets = targets.loc[common_index]
        
        # Remove last row (no target)
        features = features.iloc[:-1]
        targets = targets.iloc[:-1]
        
        # Remove NaN
        mask = ~(features.isnull().any(axis=1) | targets.isnull().any(axis=1))
        features = features[mask]
        targets = targets[mask]
        
        print(f"✅ Prepared data: {len(features)} samples, {len(features.columns)} features")
        
        return features, targets
    
    def train_models(self, symbol: str, period: str = "2y"):
        """Addestra i modelli con validazione temporale"""
        print(f"🚀 Training models for {symbol}...")
        
        # Prepare data
        features, targets = self.prepare_data(symbol, period)
        
        if len(features) < 200:
            raise ValueError("Insufficient data for training")
        
        # Create models
        self.create_models()
        
        # Scale features
        features_scaled = features.copy()
        self.scalers['robust'].fit(features)
        features_scaled = pd.DataFrame(
            self.scalers['robust'].transform(features),
            index=features.index,
            columns=features.columns
        )
        
        # Time series split for validation
        tscv = TimeSeriesSplit(n_splits=5)
        
        # Train each model
        for name, model in self.models.items():
            print(f"  📈 Training {name}...")
            
            # Cross-validation scores
            cv_scores = []
            
            for train_idx, val_idx in tscv.split(features_scaled):
                X_train = features_scaled.iloc[train_idx]
                X_val = features_scaled.iloc[val_idx]
                y_train = targets['next_return'].iloc[train_idx]
                y_val = targets['next_return'].iloc[val_idx]
                
                # Train model
                model.fit(X_train, y_train)
                
                # Validate
                y_pred = model.predict(X_val)
                score = r2_score(y_val, y_pred)
                cv_scores.append(score)
            
            avg_score = np.mean(cv_scores)
            self.validation_scores[name] = {
                'r2_scores': cv_scores,
                'mean_r2': avg_score,
                'std_r2': np.std(cv_scores)
            }
            
            print(f"    ✅ {name}: R² = {avg_score:.4f} ± {np.std(cv_scores):.4f}")
            
            # Final training on all data
            model.fit(features_scaled, targets['next_return'])
            
            # Feature importance (for tree-based models)
            if hasattr(model, 'feature_importances_'):
                importance_df = pd.DataFrame({
                    'feature': features.columns,
                    'importance': model.feature_importances_
                }).sort_values('importance', ascending=False)
                self.feature_importance[name] = importance_df
        
        self.is_trained = True
        print("🎉 Training completed!")
        
        return self.validation_scores
    
    def predict(self, symbol: str, days: int = 10) -> Dict:
        """Genera predizioni"""
        if not self.is_trained:
            raise ValueError("Models not trained yet")
        
        print(f"🔮 Generating predictions for {symbol}...")
        
        # Get recent data
        ticker = yf.Ticker(symbol)
        data = ticker.history(period="1y")
        
        # Extract features
        features = self.feature_extractor.extract_all_features(data)
        features = features.iloc[-50:]  # Last 50 days for context
        
        # Scale features
        features_scaled = pd.DataFrame(
            self.scalers['robust'].transform(features),
            index=features.index,
            columns=features.columns
        )
        
        current_price = data['Close'].iloc[-1]
        predictions = {}
        
        # Get predictions from each model
        for name, model in self.models.items():
            # Predict next return
            latest_features = features_scaled.iloc[-1:].values
            pred_return = model.predict(latest_features)[0]
            pred_price = current_price * (1 + pred_return)
            
            predictions[name] = {
                'predicted_return': pred_return,
                'predicted_price': pred_price,
                'confidence': self.validation_scores[name]['mean_r2'] if name in self.validation_scores else 0.5
            }
        
        # Ensemble prediction (weighted by validation performance)
        weights = {}
        total_weight = 0
        
        for name in predictions.keys():
            if name in self.validation_scores:
                weight = max(0.1, self.validation_scores[name]['mean_r2'])
            else:
                weight = 0.3
            weights[name] = weight
            total_weight += weight
        
        # Normalize weights
        for name in weights:
            weights[name] /= total_weight
        
        # Calculate ensemble prediction
        ensemble_return = sum(predictions[name]['predicted_return'] * weights[name] 
                            for name in predictions.keys())
        ensemble_price = current_price * (1 + ensemble_return)
        
        # Generate multi-day predictions (simplified)
        multi_day_predictions = []
        confidence_intervals = []
        
        for day in range(1, days + 1):
            # Simple drift model for multi-day
            daily_return = ensemble_return * (0.8 ** (day - 1))  # Decay factor
            pred_price = current_price * (1 + daily_return * day)
            
            # Estimate uncertainty
            model_disagreement = np.std([pred['predicted_return'] for pred in predictions.values()])
            uncertainty = model_disagreement * np.sqrt(day) * current_price
            
            multi_day_predictions.append(pred_price)
            confidence_intervals.append((pred_price - uncertainty, pred_price + uncertainty))
        
        return {
            'current_price': current_price,
            'individual_predictions': predictions,
            'ensemble_prediction': {
                'return': ensemble_return,
                'price': ensemble_price,
                'weights': weights
            },
            'multi_day_predictions': multi_day_predictions,
            'confidence_intervals': confidence_intervals,
            'data_date': data.index[-1]
        }
    
    def get_feature_analysis(self) -> Dict:
        """Analisi delle feature più importanti"""
        if not self.feature_importance:
            return {}
        
        # Combina importanza da tutti i modelli
        all_features = set()
        for model_importance in self.feature_importance.values():
            all_features.update(model_importance['feature'].tolist())
        
        combined_importance = {}
        for feature in all_features:
            importances = []
            for model_name, model_importance in self.feature_importance.items():
                feature_row = model_importance[model_importance['feature'] == feature]
                if not feature_row.empty:
                    importances.append(feature_row['importance'].iloc[0])
            
            if importances:
                combined_importance[feature] = np.mean(importances)
        
        # Sort by importance
        sorted_features = sorted(combined_importance.items(), 
                               key=lambda x: x[1], reverse=True)
        
        return {
            'top_features': sorted_features[:15],
            'individual_model_importance': self.feature_importance
        }

def analyze_stock(symbol: str) -> Tuple[str, object, object, object]:
    """Funzione principale di analisi"""
    try:
        if not symbol or len(symbol.strip()) == 0:
            return "Please enter a valid stock symbol", None, None, None
        
        symbol = symbol.upper().strip()
        
        # Initialize predictor
        predictor = FinancialPredictor()
        
        # Train models
        print(f"🎯 Starting analysis for {symbol}")
        validation_scores = predictor.train_models(symbol, period="2y")
        
        # Generate predictions
        predictions = predictor.predict(symbol, days=10)
        
        # Get feature analysis
        feature_analysis = predictor.get_feature_analysis()
        
        # Create report
        report = create_analysis_report(symbol, predictions, validation_scores, feature_analysis)
        
        # Create charts
        price_chart = create_price_chart(symbol, predictions)
        prediction_chart = create_prediction_chart(predictions)
        feature_chart = create_feature_importance_chart(feature_analysis)
        
        return report, price_chart, prediction_chart, feature_chart
        
    except Exception as e:
        error_msg = f"❌ Error analyzing {symbol}: {str(e)}"
        print(error_msg)
        import traceback
        traceback.print_exc()
        return error_msg, None, None, None

def create_analysis_report(symbol: str, predictions: Dict, validation_scores: Dict, feature_analysis: Dict) -> str:
    """Crea il report di analisi"""
    
    current_price = predictions['current_price']
    ensemble_pred = predictions['ensemble_prediction']
    individual_preds = predictions['individual_predictions']
    
    # Determine recommendation
    pred_return = ensemble_pred['return']
    if pred_return > 0.02:
        recommendation = "🟢 STRONG BUY"
    elif pred_return > 0.01:
        recommendation = "🟢 BUY"
    elif pred_return > -0.01:
        recommendation = "🟡 HOLD"
    elif pred_return > -0.02:
        recommendation = "🔴 SELL"
    else:
        recommendation = "🔴 STRONG SELL"
    
    # Model performance summary
    best_model = max(validation_scores.items(), key=lambda x: x[1]['mean_r2'])
    
    report = f"""🤖 **FINANCIAL AI PREDICTOR - {symbol}**

**📊 CURRENT STATUS**
• **Symbol:** {symbol}
• **Current Price:** ${current_price:.2f}
• **Analysis Date:** {predictions['data_date'].strftime('%Y-%m-%d %H:%M:%S')}
• **Data Quality:** ✅ High (2 years of data)

**🎯 ENSEMBLE PREDICTION**
• **Next Day Target:** ${ensemble_pred['price']:.2f}
• **Expected Return:** {pred_return*100:+.2f}%
• **Recommendation:** {recommendation}
• **Prediction Confidence:** {np.mean([p['confidence'] for p in individual_preds.values()])*100:.1f}%

**🤖 MODEL PERFORMANCE**
"""
    
    for name, scores in validation_scores.items():
        report += f"• **{name.title()}:** R² = {scores['mean_r2']:.4f} ± {scores['std_r2']:.4f}\n"
    
    report += f"""
**🏆 Best Model:** {best_model[0].title()} (R² = {best_model[1]['mean_r2']:.4f})

**📈 INDIVIDUAL MODEL PREDICTIONS**
"""
    
    for name, pred in individual_preds.items():
        weight = ensemble_pred['weights'][name]
        report += f"• **{name.title()}:** ${pred['predicted_price']:.2f} ({pred['predicted_return']*100:+.2f}%) - Weight: {weight*100:.1f}%\n"
    
    report += f"""
**🔮 MULTI-DAY FORECAST**
"""
    
    for i, (price, (low, high)) in enumerate(zip(predictions['multi_day_predictions'], 
                                                 predictions['confidence_intervals']), 1):
        report += f"• **Day {i}:** ${price:.2f} (Range: ${low:.2f} - ${high:.2f})\n"
    
    if feature_analysis and 'top_features' in feature_analysis:
        report += f"""
**🧠 TOP PREDICTIVE FEATURES**
"""
        for feature, importance in feature_analysis['top_features'][:10]:
            report += f"• **{feature}:** {importance:.4f}\n"
    
    report += f"""
**⚙️ TECHNICAL DETAILS**
• **Feature Engineering:** {len(predictor.feature_extractor.feature_names) if hasattr(predictor, 'feature_extractor') else 'N/A'} technical indicators
• **Model Architecture:** Ensemble of Random Forest + Gradient Boosting + Ridge
• **Validation Method:** Time Series Cross-Validation (5 folds)
• **Scaling:** Robust Scaler (outlier-resistant)
• **Data Source:** Yahoo Finance (real-time)

**🎯 METHODOLOGY**
• **Feature Extraction:** Technical indicators, momentum, volatility, volume analysis
• **Model Training:** Time series aware validation with walk-forward analysis
• **Ensemble Weighting:** Performance-based weighted averaging
• **Risk Management:** Confidence intervals based on model disagreement

**⚠️ DISCLAIMER**
This analysis uses machine learning models trained on historical data. Past performance does not guarantee future results. This is for educational purposes only and not financial advice. Always do your own research and consider consulting a financial advisor.

**📊 CONFIDENCE METRICS**
• **Data Sufficiency:** ✅ 2+ years of training data
• **Model Validation:** ✅ Time series cross-validation
• **Feature Quality:** ✅ {len(predictor.feature_extractor.feature_names) if hasattr(predictor, 'feature_extractor') else 'N/A'} engineered features
• **Ensemble Robustness:** ✅ Multiple model consensus
"""
    
    return report

def create_price_chart(symbol: str, predictions: Dict) -> object:
    """Crea il grafico dei prezzi con predizioni"""
    
    # Get historical data
    ticker = yf.Ticker(symbol)
    data = ticker.history(period="6mo")
    
    fig = go.Figure()
    
    # Historical prices
    fig.add_trace(go.Scatter(
        x=data.index,
        y=data['Close'],
        mode='lines',
        name='Historical Price',
        line=dict(color='blue', width=2)
    ))
    
    # Current price marker
    fig.add_trace(go.Scatter(
        x=[data.index[-1]],
        y=[predictions['current_price']],
        mode='markers',
        name='Current Price',
        marker=dict(color='red', size=10, symbol='diamond')
    ))
    
    # Predictions
    if predictions['multi_day_predictions']:
        future_dates = pd.date_range(
            start=data.index[-1] + timedelta(days=1),
            periods=len(predictions['multi_day_predictions'])
        )
        
        fig.add_trace(go.Scatter(
            x=future_dates,
            y=predictions['multi_day_predictions'],
            mode='lines+markers',
            name='AI Predictions',
            line=dict(color='red', width=3, dash='dash')
        ))
        
        # Confidence intervals
        if predictions['confidence_intervals']:
            upper_ci = [ci[1] for ci in predictions['confidence_intervals']]
            lower_ci = [ci[0] for ci in predictions['confidence_intervals']]
            
            fig.add_trace(go.Scatter(
                x=future_dates,
                y=upper_ci,
                mode='lines',
                line=dict(color='rgba(255,0,0,0)'),
                showlegend=False
            ))
            
            fig.add_trace(go.Scatter(
                x=future_dates,
                y=lower_ci,
                mode='lines',
                fill='tonexty',
                fillcolor='rgba(255,0,0,0.2)',
                line=dict(color='rgba(255,0,0,0)'),
                name='Confidence Interval'
            ))
    
    fig.update_layout(
        title=f'{symbol} - AI Price Prediction',
        xaxis_title='Date',
        yaxis_title='Price ($)',
        template='plotly_white',
        height=500,
        showlegend=True
    )
    
    return fig

def create_prediction_chart(predictions: Dict) -> object:
    """Crea il grafico delle predizioni individuali"""
    
    individual_preds = predictions['individual_predictions']
    ensemble_pred = predictions['ensemble_prediction']
    
    models = list(individual_preds.keys())
    predicted_prices = [individual_preds[model]['predicted_price'] for model in models]
    predicted_returns = [individual_preds[model]['predicted_return'] * 100 for model in models]
    weights = [ensemble_pred['weights'][model] * 100 for model in models]
    
    fig = go.Figure()
    
    # Predicted prices
    fig.add_trace(go.Bar(
        x=models,
        y=predicted_prices,
        name='Predicted Price ($)',
        marker_color='lightblue',
        yaxis='y'
    ))
    
    # Predicted returns
    fig.add_trace(go.Scatter(
        x=models,
        y=predicted_returns,
        mode='markers+lines',
        name='Predicted Return (%)',
        marker=dict(color='red', size=10),
        yaxis='y2'
    ))
    
    # Ensemble prediction line
    fig.add_hline(
        y=ensemble_pred['price'],
        line_dash="dash",
        line_color="green",
        annotation_text=f"Ensemble: ${ensemble_pred['price']:.2f}"
    )
    
    fig.update_layout(
        title='Individual Model Predictions',
        xaxis_title='Model',
        yaxis=dict(title='Predicted Price ($)', side='left'),
        yaxis2=dict(title='Predicted Return (%)', side='right', overlaying='y'),
        template='plotly_white',
        height=400
    )
    
    return fig

def create_feature_importance_chart(feature_analysis: Dict) -> object:
    """Crea il grafico dell'importanza delle feature"""
    
    if not feature_analysis or 'top_features' not in feature_analysis:
        # Empty chart
        fig = go.Figure()
        fig.add_annotation(
            text="Feature importance not available",
            xref="paper", yref="paper",
            x=0.5, y=0.5, showarrow=False,
            font=dict(size=16)
        )
        fig.update_layout(title="Feature Importance", height=400)
        return fig
    
    top_features = feature_analysis['top_features'][:15]
    features = [f[0] for f in top_features]
    importance = [f[1] for f in top_features]
    
    fig = go.Figure(go.Bar(
        x=importance,
        y=features,
        orientation='h',
        marker_color='green'
    ))
    
    fig.update_layout(
        title='Top 15 Most Important Features',
        xaxis_title='Importance Score',
        yaxis_title='Feature',
        template='plotly_white',
        height=600,
        yaxis=dict(autorange="reversed")
    )
    
    return fig

def create_interface():
    """Crea l'interfaccia Gradio"""
    
    with gr.Blocks(title="🤖 Financial AI Predictor", theme=gr.themes.Soft()) as interface:
        
        gr.Markdown("""
        # 🤖 Financial AI Predictor
        
        **Advanced Machine Learning for Stock Prediction**
        
        This system uses state-of-the-art machine learning models to predict stock prices:
        
        - 🧠 **Pre-trained Generalista Models**: Random Forest + Gradient Boosting + Ridge Regression
        - 📊 **Advanced Feature Engineering**: 50+ technical indicators and market features
        - ⏰ **Real-time Data**: Live data from Yahoo Finance
        - 🔬 **Robust Validation**: Time series cross-validation
        - 🎯 **Ensemble Predictions**: Model consensus for better accuracy
        
        **Built with scikit-learn, yfinance, and advanced feature engineering**
        """)
        
        with gr.Row():
            with gr.Column(scale=1):
                symbol_input = gr.Textbox(
                    label="📈 Stock Symbol",
                    placeholder="Enter symbol (e.g., AAPL, GOOGL, TSLA)",
                    value="AAPL"
                )
                
                analyze_btn = gr.Button(
                    "🤖 Run AI Analysis",
                    variant="primary",
                    size="lg"
                )
                
                gr.Markdown("""
                ### 🎯 How it works:
                
                **1. Data Collection**
                - Downloads 2 years of historical data
                - Real-time price and volume data
                - Technical indicators calculation
                
                **2. Feature Engineering**
                - 50+ technical indicators
                - Price momentum features
                - Volume analysis
                - Market regime detection
                
                **3. Model Training**
                - Random Forest (ensemble learning)
                - Gradient Boosting (sequential learning)
                - Ridge Regression (linear baseline)
                - Time series cross-validation
                
                **4. Ensemble Prediction**
                - Weighted model consensus
                - Confidence interval estimation
                - Multi-day forecasting
                """)
        
        with gr.Row():
            with gr.Column(scale=2):
                analysis_output = gr.Textbox(
                    label="🤖 AI Analysis Report",
                    lines=40,
                    show_copy_button=True
                )
        
        with gr.Row():
            with gr.Column():
                price_chart = gr.Plot(
                    label="📈 Price Prediction Chart"
                )
            
            with gr.Column():
                prediction_chart = gr.Plot(
                    label="🎯 Model Predictions"
                )
        
        with gr.Row():
            feature_chart = gr.Plot(
                label="🧠 Feature Importance Analysis"
            )
        
        gr.Markdown("""
        ---
        ### 🔬 Technical Details
        
        **📊 Feature Engineering Pipeline:**
        - **Price Features**: Returns, momentum, moving averages, ratios
        - **Volume Features**: Volume trends, price-volume relationships
        - **Technical Indicators**: RSI, MACD, Bollinger Bands, ADX, Stochastic
        - **Market Regime**: Volatility regimes, trend strength, gap analysis
        - **Risk Features**: Volatility ratios, drawdown indicators
        
        **🤖 Model Architecture:**
        - **Random Forest**: 200 trees, max depth 15, robust to overfitting
        - **Gradient Boosting**: 150 estimators, adaptive learning rate
        - **Ridge Regression**: L2 regularization, linear baseline model
        - **Ensemble**: Performance-weighted model averaging
        
        **🔍 Validation Strategy:**
        - **Time Series Split**: 5-fold chronological validation
        - **Walk-Forward**: Respects temporal structure of financial data
        - **Robust Scaling**: Handles outliers in financial data
        - **Feature Selection**: Automatic removal of low-quality features
        
        **📈 Prediction Pipeline:**
        1. **Real-time Data**: Fetches latest market data via yfinance
        2. **Feature Extraction**: Calculates all technical indicators
        3. **Model Inference**: Each model generates independent prediction
        4. **Ensemble**: Weighted average based on validation performance
        5. **Uncertainty**: Confidence intervals from model disagreement
        
        **⚙️ Key Advantages:**
        - **No API Keys Required**: Uses free Yahoo Finance data
        - **Real-time**: Fresh predictions on every request
        - **Robust**: Multiple models reduce overfitting risk
        - **Transparent**: Shows feature importance and model weights
        - **Educational**: Clear methodology and validation metrics
        
        **🎯 Performance Metrics:**
        - **R² Score**: Coefficient of determination (higher = better fit)
        - **Cross-Validation**: Time series aware validation
        - **Feature Importance**: Which indicators drive predictions
        - **Model Weights**: How much each model contributes
        - **Confidence Intervals**: Uncertainty quantification
        
        **📊 Supported Assets:**
        - **US Stocks**: All major exchanges (NYSE, NASDAQ)
        - **International**: Many global markets via Yahoo Finance
        - **ETFs**: Index funds and sector ETFs
        - **Crypto**: Bitcoin, Ethereum (BTC-USD, ETH-USD)
        - **Indices**: S&P 500 (^GSPC), NASDAQ (^IXIC)
        
        ### ⚠️ Important Disclaimers
        
        **🎓 Educational Purpose:**
        - This tool is designed for learning about machine learning in finance
        - Demonstrates modern ML techniques applied to financial data
        - Shows how to build robust prediction systems
        
        **📉 Financial Risk Warning:**
        - **Not Financial Advice**: Predictions are for educational purposes only
        - **Past Performance**: Historical data doesn't guarantee future results
        - **Model Limitations**: ML models can fail during market regime changes
        - **Risk Management**: Always use proper position sizing and stop losses
        - **Professional Advice**: Consult qualified financial advisors for investment decisions
        
        **🔬 Technical Limitations:**
        - **Market Efficiency**: Markets may already price in predictable patterns
        - **Black Swan Events**: Models cannot predict unprecedented events
        - **Regime Changes**: Performance may degrade during market shifts
        - **Data Quality**: Predictions depend on data quality and availability
        - **Overfitting Risk**: Models may overfit to historical patterns
        
        **🛡️ Best Practices:**
        - Use predictions as one input among many in your analysis
        - Always validate predictions against fundamental analysis
        - Consider multiple timeframes and market conditions
        - Implement proper risk management strategies
        - Continuously monitor and retrain models
        
        ### 🚀 Advanced Features
        
        **🧠 Machine Learning Pipeline:**
        ```python
        # Feature Engineering
        features = extract_technical_indicators(data)
        features = add_momentum_features(features)
        features = add_volatility_features(features)
        
        # Model Training
        models = [RandomForest(), GradientBoosting(), Ridge()]
        ensemble = train_ensemble(models, features, targets)
        
        # Prediction
        prediction = ensemble.predict(latest_features)
        confidence = calculate_uncertainty(models, prediction)
        ```
        
        **📊 Real-time Data Processing:**
        - Automatic data fetching from Yahoo Finance
        - Real-time feature calculation
        - Dynamic model updates
        - Live prediction generation
        
        **🔍 Feature Analysis:**
        - Identifies most predictive technical indicators
        - Shows feature importance across models
        - Helps understand what drives predictions
        - Guides feature engineering improvements
        
        **🎯 Ensemble Intelligence:**
        - Combines strengths of different algorithms
        - Reduces single-model bias
        - Provides uncertainty quantification
        - Improves prediction robustness
        """)
        
        # Connect the interface
        analyze_btn.click(
            fn=analyze_stock,
            inputs=[symbol_input],
            outputs=[analysis_output, price_chart, prediction_chart, feature_chart]
        )
        
        # Example stocks
        gr.Examples(
            examples=[
                ["AAPL"],   # Apple - tech giant
                ["GOOGL"],  # Google - search/AI leader  
                ["MSFT"],   # Microsoft - cloud computing
                ["TSLA"],   # Tesla - EV/energy
                ["NVDA"],   # NVIDIA - AI/semiconductors
                ["AMZN"],   # Amazon - e-commerce/cloud
                ["META"],   # Meta - social media
                ["JPM"],    # JPMorgan - banking
                ["JNJ"],    # Johnson & Johnson - healthcare
                ["V"],      # Visa - payments
                ["SPY"],    # S&P 500 ETF
                ["QQQ"],    # NASDAQ 100 ETF
                ["BTC-USD"], # Bitcoin
                ["ETH-USD"], # Ethereum
                ["^GSPC"],  # S&P 500 Index
            ],
            inputs=[symbol_input],
            label="📈 Popular Stocks & ETFs"
        )
    
    return interface

# Global predictor instance
predictor = FinancialPredictor()

def main():
    """Funzione principale"""
    print("🤖 Starting Financial AI Predictor...")
    print("📊 Checking dependencies...")
    
    # Check dependencies
    try:
        import yfinance as yf
        import ta
        from sklearn.ensemble import RandomForestRegressor
        print("✅ All dependencies available")
    except ImportError as e:
        print(f"❌ Missing dependency: {e}")
        print("Install with: pip install yfinance ta scikit-learn plotly gradio pandas numpy")
        return
    
    # Test data fetch
    try:
        print("🔍 Testing data connection...")
        test_ticker = yf.Ticker("AAPL")
        test_data = test_ticker.history(period="5d")
        if len(test_data) > 0:
            print(f"✅ Data connection OK: {len(test_data)} days of AAPL data")
        else:
            print("⚠️ Data connection: No data returned")
    except Exception as e:
        print(f"⚠️ Data connection test failed: {e}")
    
    print("=" * 60)
    print("🎯 Financial AI Predictor Features:")
    print("✅ Real-time data from Yahoo Finance")
    print("✅ Advanced feature engineering (50+ indicators)")
    print("✅ Ensemble ML models (RF + GB + Ridge)")
    print("✅ Time series cross-validation")
    print("✅ Confidence interval estimation")
    print("✅ Feature importance analysis")
    print("✅ Multi-day forecasting")
    print("=" * 60)
    
    try:
        print("🚀 Creating interface...")
        interface = create_interface()
        print("✅ Interface created successfully")
        
        print("\n🌐 Launching Financial AI Predictor...")
        print("📱 Local URL: http://localhost:7860")
        print("🌍 Public URL will be displayed below...")
        print("🤖 Ready for stock analysis!")
        print("=" * 60)
        
        interface.launch(
            server_name="0.0.0.0",
            server_port=7860,
            share=True,
            show_error=True,
            debug=False
        )
        
    except KeyboardInterrupt:
        print("\n\n🛑 Financial AI Predictor stopped by user")
        print("👋 Thanks for using the Financial AI Predictor!")
        
    except Exception as e:
        print(f"\n❌ Failed to launch: {e}")
        print("\n🔧 Troubleshooting:")
        print("1. Check if port 7860 is available")
        print("2. Install dependencies:")
        print("   pip install gradio yfinance pandas numpy plotly scikit-learn ta")
        print("3. Check internet connection")
        print("4. Try a different port: interface.launch(server_port=7861)")
        
        import traceback
        traceback.print_exc()

if __name__ == "__main__":
    main()