Apply machine learning to financial risk management  

Accurately assess financial risk levels using machine learning on a mainframe

Last updated

Machine learning is transforming all areas of business, including the way in which financial institutions and other industries are approaching tighter compliance requirements and risk management. This developer journey shows you how machine learning on IBM z/OS® is deployed for a financial risk model to determine customer credit worthiness. You’ll learn how the results are displayed using an API, enabling you to incorporate the data into business applications.

By Alexis Chretienne, Yann Kindelberger, Nora Kissari

Overview

Financial institutions need to continually weigh the risks of their transactions, and they determine their risk level through credit scoring. Leading up to the 2008-09 financial crisis, almost all large banks used credit scoring models based on statistical theories; that crisis, largely brought about by underestimating risk, proved the need for better accuracy in their scoring. The combination of increased requirements and the development of advanced new technologies has given rise to a new era: credit scoring using machine learning.

Machine Learning for IBM z/OS gives organizations the ability to quickly ingest and transform data. They can now create, deploy, and manage high quality self-learning behavioral models, using large corporate data sets residing on IBM Z. This risk assessment and management takes place securely in place and in real time, and helps financial institutions more accurately determine credit worthiness and other business needs.

In this developer journey, you will use a financial risk management model that’s been designed and deployed in a large banking system to approve or deny a loan according to input parameters. You will discover and test a financial risk management API and then create and extend a financial risk management application based on Machine Learning on z/OS. By completing the journey, you’ll discover how machine learning can be used in applications to deliver accurate financial risk management.

Flow

  1. The user calls the financial risk management API through the financial application. This API is published in a secure API Connect server, hosted in the public cloud (Bluemix).
  2. The financial risk management API calls the financial risk management system through the IBM Secure Gateway service, meaning that a Secure Gateway server has been set up in the public cloud, and a Secure Gateway client has been set up in the provider cloud, in front of the mainframe in a virtual IBM DataPower® Gateway. According to a configured access control list (ACL) file, the DataPower Gateway authorizes or denies the incoming request.
  3. If the request has been authorized by the virtual IBM DataPower Gateway, the financial risk management system (that is, the machine learning scoring service) is called through a REST/JSON interface, wherein a predictive model has been deployed. This model returns a score representing a loan approval (probability of the capacity of loan refund for a banking customer).

Components

API Connect

Create and run secure APIs and microservices.

Swagger

A framework of API developer tools for the OpenAPI Specification that enables development across the entire API lifecycle.

Secure Gateway

A service for establishing a secure, persistent connection between your environment and the cloud.

IBM DB2 for z/OS

An enterprise data server for critical business transactions and analytics.

IBM DataPower Gateway

A single multi-channel gateway that helps provide security, control, integration and optimized access to a full range of mobile, web, API, SOA, B2B, and cloud workloads.

IBM Machine Learning for z/OS

Create, deploy and manage self-learning behavioral models to extract hidden value from enterprise data securely, in place and in real time.

IBM Z Mainframe

An IBM mainframe computer in the family of z mainframe computers.

Technologies

Analytics

Finding patterns in data to derive information.

API Management

The process of creating, documenting, and making APIs available, providing access controls, and tracking statistics.

Cloud

Accessing computer and information technology resources through the Internet.

Databases

Repository for storing and managing collections of data.

Hybrid Integration

Enabling customers to draw on the capabilities of public cloud service providers while using private cloud deployment for sensitive applications and data.

Systems

Various hardware and operating systems that act as servers or hosts for applications

Related Links

learning.

https://knowledge.insead.edu/node/4837/pdf