About AI.Rating
AI.Rating is a cutting-edge platform offering credit ratings and financial insights based on advanced Machine Learning (ML) technology.
Our model is built using XGBoost technology, recognized for its predictive accuracy and efficiency.
With AI.Rating, over 39,000 public non-financial companies from 128 countries have received reliable credit ratings, revolutionizing the credit rating industry by making ratings affordable for everyone. For just 1 NZD, users can assign ratings to companies, breaking down the barriers of traditional elite rating agencies.
Our Technology
AI.Rating’s model utilizes XGBoost, which has proven its accuracy in credit analysis (link 1, link 2) and excels in programming competitions like Kaggle (more details).
This model is trained on Big Data, giving it unique advantages, such as:
- Better generalization on unseen data.
- Automatic detection of complex dependencies.
- Noise reduction and bias mitigation.
Our ML-driven model operates independently, avoiding conflicts of interest inherent in traditional sell-side rating models. The model’s transparency and reliance on financial data eliminate human bias and subjectivity.
Scientific Foundation
AI.Rating is based on the research outlined in the doctoral dissertation The Need for Disruption in the Credit Ratings Landscape: A Model for Machine Learning Computed Credit Ratings, which proposes an expert (ML-based) rating system to replace human judgment in credit rating assignments.
AI.Rating Scale and Rating Actions
The AI.Rating scale consists of 21 levels. The highest rating (maximum creditworthiness) is 1, which corresponds to “AAA” ratings from Standard & Poor’s and Fitch or “Aaa” from Moody’s. The lowest rating (maximum probability of default) is 21, which corresponds to “C” ratings from the Big Three agencies. The distribution of companies across rating categories approximately matches that of the Big Three agencies.
Along with the ratings, the rating action is also displayed on the AI.Rating website:
- I: Assignment of a rating based on the company's financial statements published on the specified date.
- RW: Revision of the company's rating due to the application of an updated rating model—on the last calendar day of each year.
- WD: Withdrawal of the company's rating because new financial statements are not available.
- DE: Withdrawal of the company's rating due to the occurrence of a default event on the specified date.
Accuracy Ratios Comparison
Below is a table comparing accuracy ratios (AR) and defaults segregation accuracy (AR’) of the AI.Rating model versus the Big Three rating agencies over the past five years (2020–2024) for different periods (1, 3, and 5 years prior to default events).
| Performance | AR 1-Year | AR'1-Year | AR 3-Year | AR'3-Year | AR 5-Year | AR'5-Year |
|---|---|---|---|---|---|---|
| AI.Rating v30_23 (All companies) | 52.61% | 49.01% | 31.50% | 30.49% | 27.57% | 24.98% |
| Big Three Average | 69.77% | 44.17% | 66.81% | 40.56% | 53.87% | 29.50% |
| AI.Rating v30_23 (Big Three-rated companies) | 80.40% | 61.78% | 68.31% | 48.33% | 62.62% | 45.57% |
Default Rates by Rating Group
The following table displays the default rates by AI.Rating groups for all companies and companies which were also rated by the Big Three agencies.
| Rating Group | All Companies (N=39,228) | S&P (N=2,745) | Moody's (N=1,963) | Fitch (N=1,183) |
|---|---|---|---|---|
| 1-3 ("AAA"-"AA") | 0.00% | 0.00% | 0.00% | 0.00% |
| 4-7 ("AA-"-"A-") | 0.05% - 0.25% | 0.00% - 0.25% | 0.00% - 0.28% | 0.00% - 0.33% |
| 8-13 ("BBB+"-"BB-") | 0.20% - 0.31% | 0.14% - 0.85% | 0.08% - 0.96% | 0.12% - 1.15% |
| 14-16 ("B+"-"B-") | 0.45% - 0.99% | 1.93% - 4.29% | 1.83% - 4.05% | 2.34% - 4.47% |
| 17-21 ("CCC+"-"C") | 0.95% - 3.60% | 6.46% - 17.03% | 7.15% - 16.14% | 8.89% - 17.17% |
Accessibility and Efficiency
The AI.Rating platform assigns ratings with exceptional speed and cost-efficiency. Our model assigns one rating in 0.018 seconds, significantly faster than traditional agencies, which can take up to a month. The entire system runs with just 1 employee and 1 computer, making the process extremely cost-effective compared to legacy systems.
How to Use AI.Rating
AI.Rating provides users with:
- Free Access: Ratings for the top 1,000 global companies are available for free.
- Paid Access: Ratings for over 38,000 companies are available through our paywall.
Custom Ratings
AI.Rating offers the ability to generate custom ratings for companies by uploading financial data. Custom ratings are useful for:
- Assessing counterparties.
- Evaluating credit risk for investments.
- Regulatory and tax purposes.