|
Abstract:
|
It all starts with a spark, an idea, and it ends in a model that goes into production; however, this is only the beginning of the journey.
As data scientists and statisticians, we are largely involved in predictive models development and we feel realized when our 'baby' graduates. We look at the KS or R2 and rejoice with the lift over an older model, the coverage on a population we never had access before, and more. Immediately questions arise from internal and external parties challenging the quality and credibility of our model. We must ensure the soundness of our work and perform a thorough validation of our inputs, modeling process and outputs. And that is not all; periodic monitoring and validation is highly advisable, if not required.
Changes in the regulatory environment, economic trends, technology, and business operations, among others, can affect how models perform and comply within the current environment. We will discuss model risk management best practices to maintain the quality of our analytics and some examples of what could happen when you turn to the dark side of the force, with a focus on the financial services industry.
|
