Factorization Machine Based Business Credit Scoring by Leveraging Internet Data

Abstract

Small business accounts for an important part in national economy. However, there are many difficulties for small business to get loan from big banks. Although big banks have begun to accept small loan over the last decades, the threshold is relatively too high. Therefore, many banks have to search help from outside vendors that provide small business credit scoring service, which in return increases their costs. Existing small business credit scoring systems ignore the Internet information, especially Web. This paper introduces a factorization machine model to predict credit scores for small firms. In this model, we combine firms’ basic information with Internet data. Experimental results show that our result is better than traditional linear regression. A demo system is given in the end.

This research was undertaken as part of Project 15BGL048, 2015AA015403, 2015BAA072 and 61303029 and supported by Hubei Key Laboratory of Transportation Internet of Things.