CSCarlos Eduardo Saavedra-Vasconez

Carlos Eduardo Saavedra-Vasconez

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Carlos Eduardo Saavedra-Vasconez has a total of 10 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Exploring Synthetic Data in Financial Analytics: A Bibliometric Review of Generation Methods, Validation Challenges, and Future Directions

    by Alexander Haro-Sarango, Diana Castillo-Martínez, Silvia Cachay-Salcedo, Jessica Saavedra-Vasconez, José Campos-Ku, Briseida Sotelo-Castro, Carlos Eduardo Saavedra-Vasconez

    Synthetic data generation has become a strategic capability in financial analytics, promising to reconcile the sector’s appetite for data-hungry machine learning with strict privacy, confidentiality, and regulatory constraints. This review provides a structured, evidence-based account of the field that is both conceptual and empirical. On the conceptual side, it explains how the principal generators work, states their objectives formally, and derives the practical consequences of each mechanism for financial data; it then organises evaluation into an operational fidelity–utility–privacy framework, extends it to incorporate the fairness and model-risk requirements imposed by regulated lending and insurance and the validation constraints that differ by data type, and offers explicit guidance on selecting a generator. On the empirical side, it maps the intellectual structure of the field through a science-mapping study of 313 Scopus-indexed documents (1997–2026) retrieved with a four-block Boolean query executed on 27 July 2026 and screened from 591 retrieved records. Performance analysis reveals sustained exponential growth, at an annual rate of 14.1% to 17.5% depending on how the endpoints are set, with 2026 covering only the first 208 days of the year, and 73.8% of output published since 2023. The corpus has accumulated 4359 citations with an h-index of 35 and pronounced inequality (Gini = 0.737; 26.2% of documents uncited), although 96.3% of the uncited documents appeared in 2025 or 2026 and the Gini falls to 0.599 once the analysis is restricted to documents with at least eighteen months of citation exposure, so these figures reflect the citation window as much as scientific impact. Co-word analysis, a strategic thematic map computed from the equivalence index, and an unsupervised clustering pipeline validated using the silhouette, Davies–Bouldin, and Calinski–Harabasz indices, together with seed, subsample, and specification stability tests, identify five overlapping and provisional research fronts: risk modelling and Monte Carlo simulation; fraud and credit-risk detection under class imbalance; privacy-preserving generation; machine-learning analytics and benchmarking; and biomedical simulation methods transferable to finance. The internal validation indices are weak in absolute terms, so these fronts are best read as an exploratory summary of a densely connected intellectual space rather than as sharply delineated lines of enquiry. We conclude that the field’s binding constraint has shifted from generation to certification: realistic financial data can now be produced, but the evidence that would establish a given synthetic dataset as simultaneously realistic, useful and safe—downstream performance on representative financial tasks, quantified disclosure risk, and robustness across market regimes—is still assembled on a study-by-study basis rather than through shared benchmarks.

  • Open Access

    Autoencoders for Financial Anomaly Detection: A Review of Architectures, Evaluation Practice and Deployment Gaps

    by Alexander Haro-Sarango, Diana Castillo-Martínez, Silvia Cachay-Salcedo, Jessica Saavedra-Vasconez, Rosa Salcedo-Dávalos, Briseida Sotelo-Castro, Carlos Saavedra-Vasconez

    Financial institutions must decide, within the milliseconds of a payment authorisation window and on evidence that is almost entirely unlabelled, whether a transaction is fraudulent: positives account for less than 0.2% of records, confirmatory labels arrive months later as chargebacks, and the generating process adapts to whatever boundary a detector exposes. Supervised classification is poorly posed under these conditions. An autoencoder is a neural network trained to reconstruct its own input through a low-dimensional bottleneck; fitted to legitimate transactions alone, it converts an intractable labelling problem into a tractable one-class fitting problem, and the reconstruction error of a new record becomes a continuous score of atypicality. This review synthesises 85 Scopus-indexed records published between 2018 and 2026. It describes eleven architectural families and their hybridisation with resampling schemes, tree ensembles, one-class classifiers, and generative models across card payments, transaction monitoring, accounting and audit records, insurance claims, anti-money laundering screening, and on-chain value transfers. It then examines how these detectors are evaluated and finds the design space broad in architecture and narrow in evaluation: threshold selection, temporally held-out validation, drift analysis, and latency reporting are absent from almost every record examined. Closing those gaps, rather than adding architectural variants, is the priority the literature indicates.