User Profile
How it works!
About Us
Welcome Page
Score Simulator
Privacy Policy
Score History
Calculate your score
Ai recommendation
Dashboard

About the Project

💭 Inspiration

Traditional credit scores exclude 45% of adults — freelancers, gig workers, students, and underbanked communities who lack formal banking history. We believe financial opportunity shouldn't depend on having a credit card when paying rent on time for years shows the same responsibility.

ScoreBridge was built to fix this inequality.

🎓 What We Learned

Technical Insights

Credit behavior patterns are complex: The dataset revealed 26+ financial features that influence creditworthiness
AI can reduce bias when trained on behavior: Our model focuses on payment patterns, not demographics
Transparency builds trust: Breaking down scores into clear components helps users understand and improve

Data Science Discovery

Working with 100,000+ credit profiles from Kaggle's Credit Score Classification dataset taught us:

Payment behavior (delays, consistency) is the strongest predictor
Credit mix and utilization matter more than absolute income
Age of credit history shows financial maturity

Key insight: Simple, explainable models build trust even if they sacrifice 2-3% accuracy.

🛠️ How We Built It

1. Data Analysis (Day 1)

Used Kaggle's Credit Score Classification dataset
Analyzed 26 financial features across 100K+ customer profiles
Identified 4 core behavioral components that drive credit reliability

2. The ScoreBridge Index Formula

We engineered features from the raw dataset into 4 interpretable components:

SBI = 0.35·P + 0.25·I + 0.20·T + 0.20·S

P = Payment Consistency
    - Delay_from_due_date
    - Num_of_Delayed_Payment
    - Payment_of_Min_Amount

I = Income Reliability
    - Annual_Income
    - Monthly_Inhand_Salary
    - Total_EMI_per_month stability

T = Transaction Patterns
    - Credit_Utilization_Ratio
    - Payment_Behaviour
    - Monthly_Balance trends

S = Savings Stability
    - Amount_invested_monthly
    - Outstanding_Debt management
    - Credit_Mix diversity

Final Score: SBI × 550 + 300 → Range [300-850]

3. Tech Stack

ML Model: Trained on Kaggle dataset, deployed via Flask API
Backend: Spring Boot for secure data handling and business logic
Frontend: React + Vite for intuitive score visualization
Database: H2 in-memory for demo (production-ready for PostgreSQL)

4. Model Training & Validation

Preprocessed 100K+ records (handled missing values, outliers)
Feature engineering to create the 4 components
Logistic regression for interpretability
Achieved 84.9% classification accuracy with full transparency

🚧 Challenges We Faced

Challenge	Solution
Messy real-world data	Cleaned 100K+ records: handled nulls, outliers, inconsistent formats
26 features → 4 components	Feature engineering to group related metrics into interpretable categories
Balancing accuracy vs explainability	Chose transparent model over black-box for trust and compliance
48-hour integration	Parallel development: ML team trained model while backend/frontend built infrastructure

Hardest part: Transforming raw credit data into a fair, explainable scoring system that doesn't perpetuate existing biases.

🎯 Outcome

What We Delivered

✅ ML model trained on 100,000+ real credit profiles
✅ 84.9% accuracy with complete score breakdown transparency
✅ Works for diverse financial profiles (traditional + alternative workers)
✅ Full-stack application: React frontend, Spring Boot backend, Flask ML service

Real-World Impact

Our model evaluates real financial behaviors instead of just credit card history:

63M+ "credit invisible" adults in US could get fair assessments
15-20% increase in qualified applicants for lenders
Instant scoring vs 30+ days for traditional credit building

🏆 Why It Matters

"A credit score should measure how you manage money, not whether you fit a 70-year-old system."

By training on comprehensive financial data and creating an explainable scoring system, ScoreBridge proves AI can increase fairness. We're not just scoring credit — we're scoring opportunity.

Built with ❤️ at HackNomics 2025