Transaction Risk Analytics

Pattern Detection Implementation

Core pattern detection algorithm with weighted risk factor analysis:

# Complex pattern detection with weighted risk analysis
def analyze_transaction_patterns(transaction, historical_data=None):
    """
    Perform multi-factor risk analysis on transaction data with weighted scoring
    
    Args:
        transaction: Current transaction data dictionary
        historical_data: Optional historical transaction data for pattern comparison
        
    Returns:
        Dictionary containing risk factors and overall risk score
    """
    risk_factors = {}
    
    # 1. Location-based risk analysis
    risk_factors['location_anomaly'] = {
        'detected': transaction['location'] in HIGH_RISK_COUNTRIES,
        'weight': 1.5,
        'description': 'Transaction from high-risk location',
        'details': f"Location: {transaction['location']}"
    }
    
    # 2. Time-based pattern analysis
    hour = transaction['timestamp'].hour
    time_anomaly = 0 <= hour <= 5  # Night hours (midnight to 5am)
    risk_factors['time_anomaly'] = {
        'detected': time_anomaly,
        'weight': 1.0,
        'description': 'Transaction during unusual hours',
        'details': f"Time: {hour}:00 hours"
    }
    
    # 3. Amount pattern analysis with multiple conditions
    amount = transaction['amount']
    amount_anomaly = (
        amount < 10 or  # Small amounts (card testing)
        amount > 500 or  # Large amounts
        (amount >= 900 and amount <= 999.99)  # Suspicious range
    )
    risk_factors['amount_anomaly'] = {
        'detected': amount_anomaly,
        'weight': 2.0,
        'description': 'Suspicious transaction amount',
        'details': f"Amount: ${amount:.2f}"
    }
    
    # 4. Combined behavior pattern analysis
    transaction_type = transaction['transaction_type']
    behavior_risk = (
        (transaction_type == 'atm' and 0 <= hour <= 5) or  # ATM at night
        (transaction_type == 'online' and amount < 10)     # Small online transaction
    )
    risk_factors['behavior_pattern'] = {
        'detected': behavior_risk,
        'weight': 1.5,
        'description': 'Suspicious behavior pattern',
        'details': f"Type: {transaction_type}, Time: {hour}:00, Amount: ${amount:.2f}"
    }
    
    # 5. Calculate overall weighted risk score
    total_weight = sum(factor['weight'] for factor in risk_factors.values())
    weighted_score = sum(
        factor['weight'] for factor in risk_factors.values() 
        if factor['detected']
    )
    
    # Normalize score to 0-1 range
    risk_score = weighted_score / total_weight if total_weight > 0 else 0
    
    # 6. Add historical pattern analysis if data is available
    if (historical_data is not None):
        # Perform additional pattern analysis with historical data
        historical_risk = analyze_historical_patterns(transaction, historical_data)
        risk_factors['historical_anomaly'] = historical_risk
        
        # Incorporate historical risk into overall score (30% weight)
        risk_score = (risk_score * 0.7) + (historical_risk['score'] * 0.3)
    
    return {
        'risk_score': risk_score,
        'risk_factors': risk_factors,
        'risk_level': get_risk_level(risk_score),
        'confidence': calculate_confidence(risk_factors)
    }