|
|
Post by dadourson on Sept 18, 2019 10:59:12 GMT
Hi,
I am starting a project for a data science course. I see a lot of you have looked into granular loan data and i would like to know which platform has the most comprehensive data available.
I am not really after estimating the performance of a loan book, but rather to replicate (and hopefully one day improve) the origination process.
As a counterexample, Mintos has a history of millions of loans but does not report many borrower features, apart from country, age, sex payment history on current loan (and the later are only available for current available loans and need to scraped from the web). It's no criticism: Mintos is not an originator so they don't own that data.
There are a lot of platform out there, any idea where to start ?
|
|
Greenwood2
Member of DD Central
Posts: 4,385
Likes: 2,784
Member is Online
|
Post by Greenwood2 on Sept 18, 2019 11:44:02 GMT
Hi, I am starting a project for a data science course. I see a lot of you have looked into granular loan data and i would like to know which platform has the most comprehensive data available. I am not really after estimating the performance of a loan book, but rather to replicate (and hopefully one day improve) the origination process. As a counterexample, Mintos has a history of millions of loans but does not report many borrower features, apart from country, age, sex payment history on current loan (and the later are only available for current available loans and need to scraped from the web). It's no criticism: Mintos is not an originator so they don't own that data. There are a lot of platform out there, any idea where to start ? I don't think any of them give much detail of borrowers. Zopa have a downloadable full loan book, which must be huge if it goes back to 2005, but I don't think it include much (or anything) about borrowers.
|
|
jlend
Member of DD Central
Posts: 1,840
Likes: 1,465
|
Post by jlend on Sept 18, 2019 14:05:01 GMT
Hi, I am starting a project for a data science course. I see a lot of you have looked into granular loan data and i would like to know which platform has the most comprehensive data available. I am not really after estimating the performance of a loan book, but rather to replicate (and hopefully one day improve) the origination process. As a counterexample, Mintos has a history of millions of loans but does not report many borrower features, apart from country, age, sex payment history on current loan (and the later are only available for current available loans and need to scraped from the web). It's no criticism: Mintos is not an originator so they don't own that data. There are a lot of platform out there, any idea where to start ? Probably best to contact platforms directly. Perhaps start with those listed on BRISMO as those have been willing to share their loan books for scrutiny. brismo.com/market-data/The very best you might get is some anonomised data where you cannot see the name of the borrower. I very much doubt you will get much else. I think in any case you are likely to need permission from platforms if you took data directly from their website for analysis.
|
|
bigfoot12
Member of DD Central
Posts: 1,817
Likes: 816
|
Post by bigfoot12 on Sept 18, 2019 14:57:14 GMT
Kaggle You may have to register to get it and some is community uploaded so watch out, but there is Kiva data and Lending Club (hundreds of MB).
|
|
|
|
Post by dan1 on Sept 18, 2019 16:51:34 GMT
You may want to get in touch with seabbs (www.samabbott.co.uk). A fellow data scientist who analysed the loan book on FC (sadly now removed as a direct download but you may like to get in touch directly), and subsequently did some consultancy work for them. |
|
