21RISK MODELING AND DATA SCIENCE

JOSHUA FRANK

Intuit Inc., Woodland Hills, CA, USA

21.1 INTRODUCTION

Risk predictive modeling is an important and growing branch of applied data science. This chapter summarizes the state of risk modeling and discusses eight key lessons learned that are applicable to risk modeling as well as to other data science endeavors. The emphasis is on building a modeling system that stresses diversity and flexibility with a “modeling ecosystem” approach. The subject matter focus is fraud and financial risk modeling for payments and credit risk applications.

21.2 WHAT IS RISK MODELING

There are many forms of business risk and therefore many definitions of risk modeling. Some types of risks are common to any large enterprise. These can include first‐party risks of hazard such as natural disasters that can damage plants and equipment, second‐party risks of hazard such as worker injuries, third‐party hazards such as liability from defective products, financial risks from sources such as foreign exchange rates and liquidity, operational risks such as the risk of labor relations issues, strategic risks from competitors and market demand, and strategic risks from regulatory/political issues as well as reputational risk [1]. There are also risks specific to the industry a company is in. For example, insurance companies have claims exposure. Lenders have risk of borrower default as well as interest rate risk for long‐term fixed rate lending. Retail merchants ...

Get Internet of Things and Data Analytics Handbook now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.