A modern, production-ready credit risk analysis system that accepts user-provided data files and performs real-time logistic regression-based risk scoring with transparent model execution traces. deployed via github pages: https://niveditha1176.github.io/Credi-AI/
- Drag & Drop Upload: Simply drag your data file onto the upload area
- Multiple Format Support: Import from .xlsx, .xls, or .csv files
- Quick Selection: Click any profile to instantly load it for analysis
- Logistic Regression Model: Industry-standard WoE (Weight of Evidence) scoring
- Transparent Calculations: Complete step-by-step model execution trace
- Real-Time Processing: Instant risk assessment for each profile
- Multi-Factor Analysis:
- CIBIL Score assessment
- Credit Utilization Rate (CUR) evaluation
- Days Past Due (DPD) tracking
- Unsecured loan count analysis
- Centered Sigmoid Graph: Beautiful probability distribution visualization
- Color-Coded Risk Levels:
- 🟢 Low Risk (PD < 25%)
- 🟡 Medium Risk (25% ≤ PD ≤ 60%)
- 🔴 High Risk (PD > 60%)
- Pipeline Flowchart: Visual representation of model execution flow
- Ledger Extract: Detailed metrics and approved financing cap
- Model Compilation Log: Complete calculation breakdown
Your data file must contain the following columns:
| Column | Type | Description |
|---|---|---|
| id | Integer | Unique customer identifier |
| name | String | Customer full name |
| income | Number | Monthly income (in currency units) |
| cibil | Integer | CIBIL/Credit score (300-900) |
| limit | Number | Total credit limit |
| outstanding | Number | Current outstanding balance |
| dpd | Integer | Days Past Due (0-180+) |
| unsecured | Integer | Number of unsecured loans |
- Click the "Import Your Data" section or drag your file
- Select your .csv, .xlsx, or .xls file
- System validates and loads all profiles
- View available profiles in the list
- Click on any profile from the loaded list, OR
- Type the Profile ID in the input field
- Profile ID will auto-complete using the datalist
- Click "Analyze Risk Profile" button
- System computes:
- WoE features for each factor
- Linear scale sum (z-score)
- Sigmoid probability (Default Probability)
- Risk classification
- Approved financing cap
- Sigmoid Graph: Shows PD on probability distribution curve
- Risk Status Badge: Visual risk category indicator
- Ledger Extract: Key metrics including:
- Credit Utilization Rate (CUR)
- Probability of Default (PD)
- Linear Scale Sum (z)
- Approved Financing Cap
- Model Log: Complete step-by-step calculations
The system uses logistic regression with Weight of Evidence (WoE) encoding:
z = -0.15 + (1.25 * WoE_CIBIL) + (1.10 * WoE_CUR) + (1.45 * WoE_DPD) + (0.85 * WoE_UNSECURED)
PD = 1 / (1 + e^z)
- CIBIL: ≥750 (0.85), 650-749 (0.15), <650 (-0.95)
- CUR: ≤30% (0.75), 30-60% (0.10), >60% (-0.80)
- DPD: 0 days (0.90), 1 day (-0.20), 2+ days (-1.10)
- Unsecured: ≤1 loan (0.40), >1 loan (-0.65)
Approved Limit = Income × Risk Factor × (1 - PD)
- Low Risk: Factor = 4.5
- Medium Risk: Factor = 1.5
- High Risk: Factor = 0
- SheetJS (XLSX): For Excel file parsing
- Vanilla JavaScript: No other frameworks required
- Modern CSS: For responsive design
- Banks & Financial Institutions: Automated credit scoring
- Lending Platforms: Quick risk assessment for loan applications
- Credit Analysis Tools: Internal risk management systems
- Financial Training: Educational demo for credit models
- Portfolio Management: Batch analysis of customer portfolios
- ✅ No data transmission to servers
- ✅ No cookies or tracking
- ✅ 100% client-side processing
- ✅ Safe for sensitive financial data
1. Export customer data from your CRM → CSV/Excel
2. Upload file to Credi AI
3. Select customer profiles
4. Get instant risk scores
5. Make informed lending decisions
Built with FinTech Grade UI | Modern Dashboard | Transparent Model Logic