Understanding pro-social landing: prediction of funding time using loan descriptions on Kiva

Kiva is an online philanthropic crowdsourcing platform that connects lenders with the borrowers in the third world. It is via the stories conveyed by the loan descriptions that the lenders get to know the borrowers and make the decision to help. In this project we study how the loan descriptions affect the funding speed of the loans. We train recurrent neural networks and long short-term memory architectures to predict whether a loan with given descriptions and characteristics will be funded quickly. It turns out that incorporating the loan descriptions as features achieves a prominent improvement over the prediction task. We also find the words and phrases associated with higher funding rates. This is crucial for Kiva to provide guidance for the description writers and remain competitive in the market.