ANOMALIE DETECTION IN THE PAYMENT SYSTEM OF ECUADOR: application of neural networks

Central banks monitor operations that are channeled through financial market infrastructures, as they are of great importance in promoting financial stability and economic growth. Transactions are a reflection of economic and commercial activity, they provide information on how banks manage liquidity. Analyzing this data with anomaly detection identifies the behavior of an unusual payment flow indicating future events and helps financial supervision to initiate timely interventions. This document makes an application of a dimension reduction method based on machine learning by neural networks (Autoencoder) for the detection of anomalies in the Interbank Payment System in Ecuador. The evaluation of the model is carried out by means of a simulation in the alteration of the flows of a bank. The results reflect that the constructed models detect the periods in which the banks presented anomalies (liquidity stress). A useful tool for monitoring operations in the payment system is proposed to identify sudden changes in flows, which could be attributed to liquidity, operational or other problems. This document is an essay of the application of the methodology constructed by Triepels et al. (2017).

• Autor principal: Rubio, Jeniffer (author)
• Altres autors: Arroyo, John (author)
• Format: article
• Idioma: spa
• Publicat: 2020
• Matèries: Anomaly Detection, Autoencoder; Neural Network; Artificial Intelligence; ACSS; Financial; Market Infraestructure; retail payments; Machine Learning; Detección de anomalías; Autoencoder; Redes Neuronales; Aprendizaje automático; Inteligencia Artificial; Sistemas de pagos; infraestructura de mercado financiero
• Accés en línia: https://estudioseconomicos.bce.fin.ec/index.php/RevistaCE/article/view/97