# ===================================================================
# MODULE 10, LESSON 8: HUMAN-CENTERED FUTURE & ETHICAL BANKING
# ===================================================================

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.metrics import confusion_matrix, classification_report
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
import warnings
warnings.filterwarnings('ignore')

print("="*70)
print("HUMAN-CENTERED BANKING – ETHICAL AI & INCLUSION")
print("="*70)

# ----------------------------------------------------------------
# PART A: SIMULATING LENDING DATA WITH POTENTIAL BIAS
# ----------------------------------------------------------------
print("\n" + "-"*60)
print("PART A: Lending Dataset with Demographic Attributes")
print("-"*60)

np.random.seed(42)

# Simulate 10,000 loan applications
n_samples = 10000
data = pd.DataFrame({
    'age': np.random.randint(18, 70, n_samples),
    'income': np.random.lognormal(10, 0.8, n_samples).round(2),  # Log-normal distribution
    'credit_score': np.random.normal(650, 100, n_samples).clip(300, 850).round(0),
    'employment_years': np.random.exponential(8, n_samples).round(1).clip(0, 45),
    'loan_amount': np.random.uniform(1000, 50000, n_samples).round(2),
    'gender': np.random.choice(['Female', 'Male', 'Non-Binary'], n_samples, p=[0.45, 0.45, 0.10]),
    'ethnicity': np.random.choice(['Group_A', 'Group_B', 'Group_C', 'Group_D'], n_samples, p=[0.4, 0.3, 0.2, 0.1]),
    'region': np.random.choice(['Urban', 'Suburban', 'Rural'], n_samples, p=[0.5, 0.3, 0.2])
})

# Introduce subtle bias: lower approval rates for certain groups
def generate_approval(row):
    base_prob = 0.7
    # Income effect
    base_prob += (row['income'] - 20000) / 100000 * 0.1
    # Credit score effect
    base_prob += (row['credit_score'] - 650) / 200 * 0.15
    # Employment effect
    base_prob += row['employment_years'] / 100
    
    # Bias: lower approval for Group_C and Group_D
    if row['ethnicity'] in ['Group_C', 'Group_D']:
        base_prob -= 0.15
    # Bias: lower approval for Rural
    if row['region'] == 'Rural':
        base_prob -= 0.08
    # Bias: slight gender bias (minor)
    if row['gender'] == 'Female':
        base_prob += 0.02  # slight positive (reverse bias for demonstration)
    
    # Clip and convert to binary
    prob = np.clip(base_prob, 0.1, 0.95)
    return 1 if np.random.random() < prob else 0

data['approved'] = data.apply(generate_approval, axis=1)

print("Loan Application Dataset (Sample):")
print(data.head(10).to_string(index=False))
print(f"\nOverall Approval Rate: {data['approved'].mean()*100:.1f}%")

# ----------------------------------------------------------------
# PART B: FAIRNESS ANALYSIS – DEMOGRAPHIC PARITY
# ----------------------------------------------------------------
print("\n" + "-"*60)
print("PART B: Fairness Analysis – Approval Rates by Demographic Group")
print("-"*60)

def calculate_approval_rates(data, group_col):
    """Calculate approval rates and metrics for fairness analysis."""
    results = data.groupby(group_col).agg(
        approval_rate=('approved', 'mean'),
        count=('approved', 'count')
    ).reset_index()
    results['approval_rate'] = results['approval_rate'] * 100
    results['disparity'] = results['approval_rate'] - results['approval_rate'].max()
    return results.sort_values('approval_rate', ascending=False)

# Approval rates by ethnicity
ethnicity_rates = calculate_approval_rates(data, 'ethnicity')
print("\nApproval Rates by Ethnicity:")
print(ethnicity_rates.to_string(index=False))

# Approval rates by region
region_rates = calculate_approval_rates(data, 'region')
print("\nApproval Rates by Region:")
print(region_rates.to_string(index=False))

# Approval rates by gender
gender_rates = calculate_approval_rates(data, 'gender')
print("\nApproval Rates by Gender:")
print(gender_rates.to_string(index=False))

# ----------------------------------------------------------------
# PART C: MODEL TRAINING WITH FAIRNESS CONSTRAINTS
# ----------------------------------------------------------------
print("\n" + "-"*60)
print("PART C: Building a Fairer AI Model with Bias Mitigation")
print("-"*60)

# Prepare features (exclude protected attributes)
features = ['age', 'income', 'credit_score', 'employment_years', 'loan_amount']
X = data[features]
y = data['approved']

# Split data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)

# Train base model (without fairness constraints)
base_model = RandomForestClassifier(n_estimators=100, random_state=42)
base_model.fit(X_train, y_train)
base_pred = base_model.predict(X_test)

# Train fair model with bias mitigation (simplified - we can reweight samples)
# Weight training samples to reduce bias
def get_sample_weights(data, y):
    """Create sample weights to balance outcomes across groups."""
    weights = np.ones(len(data))
    
    # Group C and D get higher weights to offset bias
    for idx, row in data.iterrows():
        if row['ethnicity'] in ['Group_C', 'Group_D'] and y[idx] == 0:
            weights[idx] = 1.5  # Increase weight for denied minority applicants
    return weights

sample_weights = get_sample_weights(X_train, y_train)

# Train fair model
fair_model = RandomForestClassifier(n_estimators=100, random_state=42)
fair_model.fit(X_train, y_train, sample_weight=sample_weights)
fair_pred = fair_model.predict(X_test)

# Compare models
print("\nModel Performance Comparison:")
print("\nBase Model:")
print(classification_report(y_test, base_pred, target_names=['Rejected', 'Approved']))

print("\nFair Model (with Bias Mitigation):")
print(classification_report(y_test, fair_pred, target_names=['Rejected', 'Approved']))

# Compare fairness of predictions
test_data = X_test.copy()
test_data['actual'] = y_test
test_data['base_pred'] = base_pred
test_data['fair_pred'] = fair_pred
test_data['gender'] = data.loc[X_test.index, 'gender'].values
test_data['ethnicity'] = data.loc[X_test.index, 'ethnicity'].values
test_data['region'] = data.loc[X_test.index, 'region'].values

print("\nFairness Comparison (Approval Rates by Ethnicity):")
for model, pred_col in [('Base Model', 'base_pred'), ('Fair Model', 'fair_pred')]:
    rates = test_data.groupby('ethnicity')[pred_col].mean() * 100
    print(f"\n{model}:")
    for eth, rate in rates.items():
        print(f"  {eth}: {rate:.1f}%")

# ----------------------------------------------------------------
# PART D: ALTERNATIVE CREDIT SCORING FOR FINANCIAL INCLUSION
# ----------------------------------------------------------------
print("\n" + "-"*60)
print("PART D: Alternative Credit Scoring using Non-Traditional Data")
print("-"*60)

# Simulate data for unbanked population (no credit score)
unbanked_data = pd.DataFrame({
    'customer_id': [f'U{str(i).zfill(4)}' for i in range(1, 501)],
    'mobile_usage_score': np.random.uniform(0.2, 0.95, 500),  # Proxy for stability
    'utility_payment_regularity': np.random.uniform(0.1, 0.98, 500),
    'social_network_score': np.random.uniform(0.3, 0.9, 500),
    'income_proxy': np.random.lognormal(8, 0.5, 500),  # Estimated income
    'age': np.random.randint(20, 60, 500),
    'employment_status': np.random.choice(['Employed', 'Self-Employed', 'Unemployed', 'Retired'], 500)
})

# Calculate composite alternative credit score
unbanked_data['alt_credit_score'] = (
    0.30 * unbanked_data['mobile_usage_score'] +
    0.25 * unbanked_data['utility_payment_regularity'] +
    0.20 * unbanked_data['social_network_score'] +
    0.15 * (unbanked_data['income_proxy'] / unbanked_data['income_proxy'].max()) +
    0.10 * (unbanked_data['age'] / 60)
) * 850

# Determine eligibility
unbanked_data['loan_eligible'] = unbanked_data['alt_credit_score'] > 550

print("Alternative Credit Scoring for Unbanked Population:")
print(f"Total Unbanked Applicants: {len(unbanked_data)}")
print(f"Eligible for Loans: {unbanked_data['loan_eligible'].sum()}")
print(f"Eligibility Rate: {unbanked_data['loan_eligible'].mean()*100:.1f}%")

print("\nSample Alternative Credit Scores:")
print(unbanked_data.head(10)[['customer_id', 'alt_credit_score', 'loan_eligible']].to_string(index=False))

# ----------------------------------------------------------------
# PART E: COMPREHENSIVE FUTURE STRATEGY DEVELOPMENT
# ----------------------------------------------------------------
print("\n" + "="*70)
print("PART E: Comprehensive Future Strategy for a Digital Bank")
print("="*70)

future_strategy = {
    "Vision": "To be the most trusted, inclusive, and innovative digital bank, empowering customers to achieve financial well-being in a sustainable and ethical manner.",
    
    "Mission": "Leverage cutting-edge technology and human-centered design to deliver personalized, transparent, and accessible financial services that adapt to the evolving needs of our customers and communities.",
    
    "Strategic Pillars": {
        "1. Customer-Centricity": {
            "Focus": "Deliver personalized experiences that anticipate customer needs.",
            "Initiatives": [
                "Implement AI-driven hyper-personalization for all products.",
                "Design inclusive products for underserved segments.",
                "Build a seamless omnichannel experience."
            ],
            "KPIs": ["NPS > 80", "Customer Satisfaction > 95%", "Inclusion Index > 0.8"]
        },
        "2. Technological Innovation": {
            "Focus": "Build a resilient, future-proof technology infrastructure.",
            "Initiatives": [
                "Migrate to cloud-native microservices architecture.",
                "Implement quantum-safe cryptography by 2030.",
                "Deploy autonomous AI agents for operations."
            ],
            "KPIs": ["100% Cloud Migration", "PQC Integration 100%", "Automation Rate > 80%"]
        },
        "3. Ecosystem & Partnerships": {
            "Focus": "Create a vibrant ecosystem of financial and non-financial services.",
            "Initiatives": [
                "Launch Banking-as-a-Platform with 100+ partners.",
                "Integrate embedded finance into third-party platforms.",
                "Develop open APIs with robust developer experience."
            ],
            "KPIs": ["100+ Active Partners", "API Calls > 1M/Month", "Ecosystem Revenue > $100M"]
        },
        "4. Sustainability & Ethics": {
            "Focus": "Lead in sustainable finance and ethical AI.",
            "Initiatives": [
                "Achieve net-zero financed emissions by 2050.",
                "Implement AI fairness audits for all models.",
                "Develop green lending products with ESG incentives."
            ],
            "KPIs": ["Net-Zero Target 2050", "Model Fairness Score > 0.9", "ESG Portfolio > 50%"]
        },
        "5. Talent & Culture": {
            "Focus": "Develop a future-ready workforce with a purpose-driven culture.",
            "Initiatives": [
                "Continuous learning and development programs.",
                "Build diverse and inclusive leadership teams.",
                "Foster a culture of innovation and experimentation."
            ],
            "KPIs": ["Learning Hours > 50/Year", "Diversity Index > 0.35", "Innovation Index > 4.5"]
        }
    }
}

print("🧭 Future Strategy for the Digital Bank:\n")
print(f"Vision: {future_strategy['Vision']}\n")
print(f"Mission: {future_strategy['Mission']}\n")

print("Strategic Pillars:")
for pillar, details in future_strategy['Strategic Pillars'].items():
    print(f"\n  {pillar}:")
    print(f"    Focus: {details['Focus']}")
    print("    Initiatives:")
    for initiative in details['Initiatives']:
        print(f"      • {initiative}")
    print(f"    KPIs: {', '.join(details['KPIs'])}")

# ----------------------------------------------------------------
# SECTION 6: FINAL SUMMARY AND REFLECTION
# ----------------------------------------------------------------
print("\n" + "="*70)
print("LESSON 8 SUMMARY – MODULE 10 COMPLETE")
print("="*70)

print("""
Human-Centered & Ethical Banking – Key Takeaways:

1. Human-Centered Banking prioritizes empathy, simplicity, inclusion, transparency, and trust.
2. Ethical AI requires fairness audits, explainability, accountability, and privacy protection.
3. Inclusive Banking uses alternative credit scoring to serve the unbanked and underbanked.
4. We demonstrated fairness analysis and bias mitigation using Python.
5. Alternative data (mobile usage, utility payments) enables financial inclusion.

Final Recommendations for the Future Digital Bank:
  ✓ Invest in ethical AI frameworks and regular bias audits.
  ✓ Design products for inclusion – serve the underserved.
  ✓ Build ecosystems, not just products – embrace platform economics.
  ✓ Prepare for quantum computing with PQC migration plans.
  ✓ Embed ESG into core banking strategy, not just reporting.
  ✓ Foster a culture of innovation, continuous learning, and purpose.

REFLECTION QUESTIONS FOR YOUR JOURNEY:
  1. What is your personal vision for the future of digital banking?
  2. How will you ensure that technology serves humanity, not the other way around?
  3. What role will you play in building an inclusive and sustainable financial system?
  4. What actions will you take to drive ethical AI in your organization?
  5. How will you measure your impact on customers and communities?