If you want to discover the latest ideas in credit rating – an area that is growing faster than many people think – the best place is the annual credit assessment and control conference. Edinburgh to our own FICO World ). At this year's conference, I started thinking about what has changed in the world of credit rating since my first visit to the conference nearly 20 years ago. My phone has certainly become smarter by then – while is smarter in credit rating?
With about 70 presentations, the key questions remain the same:
Which data are available and useful?
How can I get the best intelligence from the data?
How can I act intelligence?
How am I compliant (19459007) to more and more regulations?
Regarding the data the answer is growing! We all know that the digital age creates vast amounts of data that exponentially increase. Compared with 20 years ago, many new and "alternative" data sources are now being used to improve the creditworthiness of certain types of consumers, including psychometric data, telecom data, social network data, and data from the Internet. transactions.
In terms of transformation of these data into intelligence artificial intelligence (AI) and in particular the machine learning algorithms are studied and used on a larger scale. With ever-increasing volume and a greater variety of data associated with significantly increased computing power, machine-learning approaches to derive value from data are proving increasingly useful.
This same processing power for the development of machine learning models also helps in the ease of deployment of these types of models, which has been inconvenient in the past in terms of speed of deployment and speed of deployment. # 39; execution.
The evolution of predictive analytics (models that order by a single result of interest) to prescriptive analytics (also called optimization of decision, identifying the best action ] to take into account if one considers several metrics or dimensions of results) significantly improve the operational results of decisions throughout the life cycle of credit, from creation to collection and recovery. Prescriptive analysis makes it possible to make better decisions by taking into account multiple (often conflicting) objectives – for example, increasing acceptance rates while controlling losses.
Since the economic crisis, there has also been a focus on modeling stressful situations and how these tensions affect both the likely performance of individual consumers as well as the overall portfolios. Predictive models help lenders comply with regulations such as Basel and IFRS 9. As compliance is acquired and maintained, these same models are used to generate business value through a better understanding of portfolios. inputs to both hypothetical scenario analysis and decision optimization capabilities.
FICO scientists and data experts, all of whom have blogged here, have presented no less than five sessions in Edinburgh this year on hot topics related to these areas I have described.
Gerald has already written an article on his lecture, on Risk Analysis to Test Individual Consumers – we will share more ideas on our topics here, and will bring to light new trends in credit rating.