MIDS Capstone Project Fall 2016

OptiLender: P2P Portfolio Optimization with Machine Learning

P2P lending is a huge - the two leading platforms, Prosper and Lending Club, together originated $13 billion loans in 2015. For investors, it offers great profit potentials, with tremendous difficulty. In addition to thousands of different loan opportunities that emerge daily, an investor has construct a portfolio based on hundreds of attributes attached to each loan, making this an almost impossible task.

OptiLender uses Machine Learning to recommend portfolios that consistently outperforms the benchmark. We provide outperformance, transparency, and simplicity to P2P investors through:

  • Reducing the number of variables from 500 to 10
  • Outsample model validation going back to 2010
  • A decision tree model that is easy to understand
  • Intuitive and interactive frontend solutions to visualize actual model performance

With OptiLender, an investor can spend far less time on reading pages and pages of loan opportunities and understanding the various terms, and focus more on constructing a portfolio that one understands and is comfortable with.

Course

Data Science W210. Capstone - Lecture - 1 - 2016

Last updated:

March 30, 2017