A carregar...
Publicação
Application of Predicted Models in Debt Management: Developing a Machine Learning Algorithm to Predict Customer Risk at EDP Comercial
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.14 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This report is a result of a nine-month internship at EDP Comercial where the main project of research was the application of artificial intelligence tools in the field of debt management. Debt management involves a set of strategies and processes aimed at reducing or eliminating debt and the use of artificial intelligence has shown great potential to optimize these processes and minimize the risk of debt for individuals and organizations.
In terms of monitoring and controlling the creditworthiness and quality of clients, debt management has mainly been responsive and reactive, attempting to recover losses after a client has become delinquent. There is a gap in the knowledge of how to proactively identify at-risk accounts before they fall behind on payments.
To avoid the constant reactive response in the field, it was developed a machine-learning algorithm that predicts the risk of a client becoming in debt by analyzing their scorecard, which measures the quality of a client based on their infringement history.
After preprocessing the data, XGBoost was implemented to a dataset of 3M customers with at least one active contract on EDP, on electricity or gas. Hyperparameter tuning was performed on the model to reach an F1 score of 0.7850 on the training set and 0.7835 on the test set. The results were discussed and based on those, recommendations and improvements were also identified.
Descrição
Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science
Palavras-chave
Debt Management Machine Learning Scorecard Prediction Artificial intelligence