Welcome to the first month roundup starting in February 2018. Let’s see the major points to report:
Bondora Loan offers
There is a significant change in the loan offering by Bondora since beginning of this year. Started already in January but continued
now also in February. In all the past years offerings of HR loan took a share of ~ 25%. Now it dropped to 8.7%. Here
an overview of the new risk share of loan offerings
| Risk Class | AA | A | B | C | D | E | F | HR |
|---|---|---|---|---|---|---|---|---|
| 2018 | 8.6 | 8.1 | 17.0 | 17.7 | 16.5 | 13.7 | 9.7 | 8.7 |
| 2017 | 4.8 | 5.8 | 15.1 | 17.2 | 13.5 | 10.8 | 6.9 | 25.9 |
| Difference | + 3.8 | + 2.3 | + 1.9 | + 0.5 | + 3.0 | + 2.9 | + 2.8 | – 17.2 |
Artificial Intelligence
Also the Artificial Intelligence window is moved one month forward. Loans from March 2013 till February 2017 are taken into account.
What’s the “Artificial Intelligence window”?
The Window covers a certain time frame and is moved each month one month forward. So loans from 2013 are dropping out this year while loans from 2017 are coming in. Taking newest loans into the calculation ensures that the AI is constantly growing. While 2013 there are 1844 public loans in the stats 2017 has 16858 public loans! So do the math => each month 150 loans from 2013 are dropping out while 1400 loans come in. So each month the training size of the AI is raised by 1250 loans!
Why February 2017?
Only after 12 months the AI can calculate if this is a good or bad loan! Judgement is too early if e.g. after 3 months of payment the loan isn’t defaulted. So only mature loans are taken into account!
The loan default statistic moved one month forward and contains now the loans till end of February 2018. The new
year hasn’t any defaults so far, no surprise because this year till that date just 59 days old. Next month I’m pretty
sure the first defaults will show up!
For 2017 the curve for overall defaults begins to rise. You cannot see it visually very well but the first values
of “Defaults after 12 months” can be read in the values.
