Research on fraud detection and decision systems.

Current direction

My doctoral work focuses on fraud detection in payment systems, with particular attention to false positives, false negatives, temporal validity, and the economic impact of decision errors. The central problem is not simply model performance. It is whether a fraud control system improves net outcomes under real operational constraints.

This includes evaluation design, threshold policy, friction cost, governance, and the translation of technical results into better decision quality.

Core themes

Applied research framework

CyberAntifraud is the main public project linked to this research line. It brings together evaluation protocols, cost and friction logic, model documentation, and governance oriented artefacts for acquiring fraud detection. The project is designed to sit between academic rigour and operational relevance.

It is not a generic AI site. It is a structured fraud and decision infrastructure focused on clarity, reproducibility, and deployable judgement.

Project link

cyberantifraud.com

Research protocols, public artefacts, and method facing pages.

Fraud Paper Analyzer

Tool concept for extracting datasets, metrics, validation strategies, and methodological gaps from fraud detection papers.

Imbalanced Learning Evaluator

Tool concept for assessing minority class handling, metric adequacy, threshold logic, and cost sensitivity in published models.

Economic Fraud Impact Estimator

Tool concept for modelling the operational and economic consequences of false positives, false negatives, and control thresholds.

GitHub

github.com/saramago

Repositories, prototypes, and public technical artefacts.

LinkedIn

linkedin.com/in/saramago

Professional profile and public positioning.

Research outputs

Papers, review materials, project artefacts, and linked methodological outputs will be added progressively as public research deliverables become available.