Statistical Profiling of Banking Data for Anomaly Identification:An Applied Study Using a Synthetic Banking Database
Qubahan EcoLead Journal · 30.06.2022 · 1 (2)
DOI: 10.48161/qelj.v1n2a3 ↗Kurz gefasst
Fachartikel in der Zeitschrift Qubahan EcoLead Journal, online erschienen am 30.06.2022. Verfasst von Djina Ivanovic und 1 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
This study examines whether statistical analysis can meaningfully distinguish anomalous from non-anomalous records in an openly available banking database. The dataset contains 5,000 observations and 40 variables covering customer, account, transaction, loan, credit-card, and feedback information. For analysis, the original anomaly label was recoded so that 300 records (6.0%) were treated as anomalous and 4,700 records (94.0%) as normal. The empirical design combines descriptive statistics, Mann-Whitney U tests, chi-square association tests, and a logistic-regression benchmark on a hold-out sample. The results show that major numerical variables do not differ significantly between the two classes, while only account type and resolution status exhibit statistically significant but substantively small associations. The benchmark classifier performs weakly (ROC-AUC = 0.490; average precision = 0.062), indicating that the available variables provide limited discriminatory signal. The study therefore contributes less as an operational fraud-detection model and more as a transparent methodological illustration of how class imbalance, weak labels, and synthetic feature spaces can constrain inference. To strengthen the manuscript for journal submission, the present version expands the literature review with pre-2021 sources, introduces a comparison matrix against prior studies, and adds statistical tables and figures derived from the uploaded dataset.
Zitieren
APA
Ivanovic, D., & Khoza, E. (2022). Statistical Profiling of Banking Data for Anomaly Identification:An Applied Study Using a Synthetic Banking Database. Qubahan EcoLead Journal, 1(2). https://doi.org/10.48161/qelj.v1n2a3
Vancouver
Ivanovic D, Khoza E. Statistical Profiling of Banking Data for Anomaly Identification:An Applied Study Using a Synthetic Banking Database. Qubahan EcoLead Journal. 2022;1(2). doi:10.48161/qelj.v1n2a3
BibTeX
@article{Ivanovic2022,
author = {Ivanovic, Djina and Khoza, Evans},
title = {Statistical Profiling of Banking Data for Anomaly Identification:An Applied Study Using a Synthetic Banking Database},
journal = {Qubahan EcoLead Journal},
year = {2022},
volume = {1},
number = {2},
doi = {10.48161/qelj.v1n2a3}
}
EUVIMED