Transform Regional Bank's $85 Million Loss into Profit
📋 Mission Briefing
Situation Critical: Regional Bank Corporation is hemorrhaging money. Their loan approval system - based on simple credit score cutoffs - approved thousands of loans that defaulted (costing $50M) while rejecting profitable customers (losing $35M in potential revenue).
Your Mission: Build an intelligent decision tree system that learns from 5 years of loan data to make smarter approval decisions. Save the bank from collapse while maintaining regulatory compliance.
Success Criteria: Reduce losses by at least $40M annually while maintaining 60%+ approval rate.
Mission Progress
0%
Ready to begin your mission...
💰 Financial Impact Dashboard
Current Annual Loss-$85M
Potential Savings$0
Approval Rate72%
Default Rate18%
🎯 Challenge 1: Analyze the Current Crisis
First, let's understand why the current system is failing. The bank uses a simple rule: approve if credit score > 650.
Current System Performance
📊 Looking at the data above, what's the main problem with using only credit score?
💡 Insight: Credit score alone misses crucial patterns! Notice how many high-income, low-debt customers with scores of 640-650 successfully repay loans, while some with 700+ scores but high debt ratios default. We need to consider multiple factors simultaneously!
🌳 Challenge 2: Build Your Decision Tree
Now it's time to build a smarter system. Adjust the tree parameters to find the optimal configuration.
Tree Configuration
Tree Depth:3
Min Samples per Split:20
Feature Importance Threshold:0.05
💡 Pro Tip: Start with depth 4-5 for good balance. Too shallow = underfitting (misses patterns). Too deep = overfitting (memorizes training data). Watch how financial impact changes with depth!
🔬 Challenge 3: Test on Real Applications
Your tree is trained! Now test it on actual loan applications to see how it performs.
✅ Reduced annual losses from $85M to $33M (61% improvement)
✅ Increased profitable loan approvals by 15%
✅ Reduced default rate from 18% to 11%
✅ Maintained 65% approval rate (above 60% target)
✅ Achieved 87% model accuracy
✅ 100% decisions are explainable for regulatory compliance
Business Impact:
Your decision tree system will save Regional Bank $52 million annually while improving customer satisfaction through faster, more consistent decisions. The model processes applications in under 50ms, replacing the previous 3-day manual review process.