Prospects and Risks of Algorithmic Lenders Using Big Data (II)

Algorithmic Lender 2.0 is based on learning algorithm. The lender not only increases the amount of data input, but also analyzes the data through learning algorithm. It can learn how to determine its own credibility, including the priority and weight of factors of credit score. However, Algorithmic Lender 2.0 is often thought as a "black box," through which programmers can see input data and results, but not the process. If errors occur in the learning algorithm, the programmer mostly can't check the instructions the algorithm follows as he would with a traditional algorithm.

Specifically for credit scores, Algorithmic Lender 2.0 can analyze big data sets, including all data about people who have previously applied to companies for credit, whether they received credit and whether they repaid loans. It can also further identify factors and weights of factors of credit score. With a big sample of borrowers accumulated with details (including repayment histories and so on), the learning algorithm can predict the likelihood of future borrower repayments.