Crowdminr.io – Invest Smarter in Peer-to-Peer Loans



Crowdminr.io – Invest Smarter in Peer-to-Peer Loans

0 0


about

A mini-site describing crowdminr.io, built with reveal, yeoman, bower, d3, coffeescript and MathJax

On Github crowdminr / about

Crowdminr.io

Invest Smarter in Peer-to-Peer Loans

Matt Gibb - DPhil, University of Oxford

$249 Million

In the last 30 days

Proportional Hazard Model

A borrower i has frailty h(Xi)=
Hazard μ(t)
Default probability f(t)
Net Present Payment Value
Net Present Cumulative Value

Return Distribution

How are the interest rates predicted to take account of the risks? Random forests are used to estimate the frailty term for a proportional hazards model. This is based on hundreds of features of each loan, which you can think of as how generally broke the borrower is. A global 'hazard rate' is calculated using survival analysis, specifically Kaplan-Meier estimation, which in each month gives the chance of a single loan defaulting, and is multiplied by the frailty term. From here you can calculate the probabilities of default in each month. Since money is worth less in the future than now, each payment is discounted by its Net Present Value Summing all these values multiplied by their probabilities gives an expected return for each loan.
Your browser is old and doesn't support html5 video. Try Google Chrome.