During the week I share the latest p2p lending news on Twitter as it happens. Then every Saturday I take the most interesting news items and blog posts from the past week and share them here.
Lending Club Courts Small Banks as Personal-Loan Partners from American Banker – The most important news story of the week was clearly this front page article about Lending Club in the leading publication devoted to banking.
Investment Timing for Prosper Loans from Orchard – Fascinating post showing how fast loans are being funded at Prosper.
P2P lending sites facing bankruptcy in China from Want China Times – Some challenges appearing in the Chinese p2p lending market with dozens of platforms going bankrupt recently.
Big Data & P2P Lending: Don’t Be Misled from P2P Lending Advice – Why you need to be careful when doing statistical analysis on p2p lending data.
Lending Club hits $3 billion milestone from San Francisco Business Times – The recent Lending Club milestone caught the attention of Mark Calvey who has been following this industry for many years.
Third Party Automated Investing Tool for Lending Club from Peer & Social Lending – Q&A with one of the founders of LendingRobot, a new tool for investors.
Is Peer-to-Peer Lending Over? from The College Investor – As regular Lend Academy readers know p2p lending has been evolving quickly this year.
Prosper Review: 5 Steps (and 4 Days) for an Unsecured Personal Loan from LendingMemo – Simon Cunningham recently applied for a loan at Prosper and he writes about the experience in great detail.
The Cool Kids of Finance Join the Peer-to-Peer Bandwagon from American Banker – Slide show featuring the many big names now involved with p2p lending.
From the Lend Academy Forum
The Lend Academy forum is where investors go to discuss p2p lending. Below are some topics that were being discussed this week.
New LC Prime – There was a lot of discussion this week about the new PRIME offering from Lending Club.
New Front Page Pie Graph? – Discussing the differences between credit card refinancing, debt consolidation and credit card payoff.
Notified borrower of potential legal action – Yes, sometimes delinquent borrowers are sued.
Just had to comment on the Big Data article.
Disclaimer: I admit I am biased towards machine learning. I also believe that ML may not be the best solution. It can however be used to complement industry experience and knowledge.
Here are my thoughts regarding the article:
Factor 1) Agreed; it is reflective of the past. But how else would one make an informed decision?
Factor 2) Yes, more loans are being issued now than ever before. Naturally, that means there are more young notes than mature. However, that doesn’t discount the information gleaned from either. There is plenty of valuable information from current notes as there are ones to term.
Factor 3) Proper ML models do not isolate covariates. Rather they are in part designed to identify patterns and relationships among each other. A side note, there are 1R classification models that perform well for many cases.
Also, the author mentioned the redundancy of certain factors (income was mentioned) when filtering by credit score. FICO credit scores are primarily correlated with payment history and level of debt. TTBOMK, Income is not considered in the FICO calculation. Also FICO may not necessarily provide a complete credit risk assessment for too many reasons to list here.
Factor 4) This is why having understanding and experience is important when measuring risk. Correlation isn’t causation is similar to the study of intelligence to shoe size (google it if you don’t know the study). Bottom line, with any ML models, you need to have an understanding of what it is you are trying to learn. Then, factor that knowledge into the results of your big data analysis. There are many ways to do this including a base loan selection filter/strategy.
Overall, I believe the author made some valid points. It is indeed wise to be cautious of using BDA in isolation. However, there were some glaring issues that could potentially mislead readers.
PS. I posted my comments here because I don’t believe the author’s comment system is public.
Thanks for the detailed response. The reason I included the article is to remind investors that just doing historical analysis on Nickel Steamroller is not enough. You need to look at interrelated factors – as you have pointed out very well in your comments.
Peter, Thanks for sharing the article and completely agree with you. I enjoy reading your weekly posts, and really appreciate what you do for the P2P community.