Big Data

Finding hidden patterns and relationships in credit bureau data is a key to building new and advanced statistical indicators that cover various areas, from the creditworthiness assessment of subjects with no or thin credit history to anti-fraud scores, and collection. CRIF uses machine-learning techniques to exploit the value of its credit bureau data and to provide innovative solutions to problems that have so far been left unanswered.

As an example, with a strong focus on financial inclusion, CRIF has developed a machine-learning based indicator that is able to score a subject (individual, sole trader, SME) even if not present in the credit bureau database. Based on the information that is usually provided along with the application data, as well as using technical enquiry information, open data, internally calculated indicators and derived variables, the no hit score is able to return a statistical indicator of the likelihood that a customer will default on a payment over the next year, despite the absence of credit bureau data.

With a strong mix of proprietary data, dedicated Hadoop-based big data infrastructure, data analysis skills, compliance and data protection expertise, CRIF is very well positioned to continue exploiting the value of its credit bureau data, mixing machine-learning techniques with well-established standard statistical approaches.