Title: An automatic and personalized strategy for scoring, tracking and monitoring credit risk and collections using machine learning and artificial intelligence methods

Abstract:

A credit and collections strategy aims to assess the risk of a person or company in terms of granting credit and collecting reimbursements. A large number of research is primarily based on the combination and construction methods of credit scoring models. But the credit and collection strategy cannot be limited only to Credit Scoring but goes further through the management of models through their follow-up, monitoring, validation and calibration. In this sense, firstly, different types, sources, quality and quantity of credit and collection data require different model specifications and how to automatically search and construct credit scoring models according to the data available has become essential. Second, once a model is specified, how to automatically update or calibrate it once it no longer works well in the target population has become a key to optimizing time, resources, risk-return and generating correct predictions on time. Finally, how to automatically implement the models and its actualizations in real time has become necessary to ensure business agility. Those three essentials are the main concern of this proposal. In response to the current challenges for credit scoring research and development, we proposed an automatic and personalized strategy for scoring, tracking and monitoring credit risk and collections, designed a credit and collection assessment platform which includes an analytic engine, a monitoring module, a web service module, and finally the user and model management module. The analytical engine includes data import, processing, cleaning and treatment of large volumes of information through data mining, identification, creation and transformation of variables, feature selection, segmentation, classification model automatic search, parameter estimation and classification output. For this module we further incorporate automatic search ways to reduce manual interaction. The monitoring module includes analysis of discrimination, stability, concentration, calibration and risk determinants. In the web service, solutions are implemented instantly from anywhere in the world through the cloud. Lastly, the platform organizes, produces, maintains and manages the analytical models. We utilize public and self-owned credit data sets to conduct experiments and compare them with the latest credit assessment methods. Extensive experiments have demonstrated that our proposal achieves noticeable performance improvements. In addition, the findings show that our platform balances automation and accuracy, which can be implemented to the financial industry to enhance the reliability of credit and collection assessments.

Biography:

Adriana Uquillas has a PhD in Statistics from the University of São Paulo (Brazil). She is also a Mathematical Engineer from the Escuela Politécnica Nacional (Ecuador) and Bachelor in Classical Music from the Latin American Institute of Sacred Music (Ecuador). Adriana was Research and Development Manager, Manager of Credit Models and Credit Risk Manager and Senior Manager at Itaú Unibanco Bank, at Banco Múltiplo HSBC and at Banco Bradesco. Since 2016, she is co-founder and president of EstocasticAnalytics S.A company, which provides analytical services that contribute to business innovation. In the academic field, Adriana is a full professor in the Mathematics Department of the Escuela Politécnica Nacional. She has fourteen years of experience in the financial sector, dedicated to Research and Development in the banking sector. Also, she has collaborated in the Scientific Division of Energy Analysis and Development of the University of São Paulo and in the Machine Learning and Computer Vision research group of Escuela Politécnica Nacional. Currently, she is a member of the Integrative Ecology Research Group and collaborates in the Multidisciplinary Research Group for Public Policy of the Escuela Politécnica Nacional.