from tabpfn import TabPFNClassifier
import pandas as pd

# Customer data features
features = [
    'days_since_last_purchase',
    'total_purchases',
    'avg_order_value',
    'support_tickets',
    'payment_method',
    'subscription_tier'
]

# Load customer data
customer_data = pd.read_csv("customer_data.csv")
X = customer_data[features]
y = customer_data['churned']  # Binary: 0 or 1

# Predict churn probability
classifier = TabPFNClassifier()
churn_probability = classifier.predict_proba(X)[:, 1]

# Identify high-risk customers
high_risk_customers = customer_data[churn_probability > 0.7]

# Transaction features
transaction_features = [
    'amount',
    'merchant_category',
    'time_of_day',
    'day_of_week',
    'location_distance',
    'device_type',
    'previous_fraud_history'
]

# Real-time fraud detection
def detect_fraud(transaction_data):
    classifier = TabPFNClassifier()
    fraud_score = classifier.predict_proba(transaction_data)[:, 1]
    
    if fraud_score > 0.8:
        return "HIGH_RISK"
    elif fraud_score > 0.5:
        return "MEDIUM_RISK"
    else:
        return "LOW_RISK"

# Patient features
patient_features = [
    'age',
    'gender',
    'blood_pressure',
    'cholesterol',
    'glucose_level',
    'family_history',
    'lifestyle_factors'
]

# Disease prediction
def predict_disease_risk(patient_data):
    classifier = TabPFNClassifier()
    risk_probability = classifier.predict_proba(patient_data)[:, 1]
    
    risk_level = "Low" if risk_probability < 0.3 else \
                "Medium" if risk_probability < 0.7 else "High"
    
    return {
        'risk_level': risk_level,
        'probability': risk_probability[0],
        'recommendations': get_recommendations(risk_probability[0])
    }

# Product quality features
quality_features = [
    'dimension_1',
    'dimension_2',
    'weight',
    'surface_finish',
    'material_hardness',
    'temperature',
    'pressure'
]

# Quality classification
def classify_product_quality(product_data):
    classifier = TabPFNClassifier()
    quality_score = classifier.predict_proba(product_data)[:, 1]
    
    if quality_score > 0.8:
        return "PASS"
    elif quality_score > 0.5:
        return "REVIEW"
    else:
        return "FAIL"