Attention-Driven Dual-Level Cost-Sensitive Stacking for Financial Statement Fraud Detection
Bibliographic record
Abstract
Detecting fraud in financial statements is essential for maintaining transparency and accountability. In this study, we propose a novel dual-level cost-sensitive stacking ensemble framework for financial fraud detection, integrating a range of traditional classifiers with a state-of-the-art attention-based neural model. Our approach applies cost weighting at both the base and meta levels to amplify minority-class detection and pairs it with SMOTE-based oversampling. Additionally, we combine heterogeneous financial ratios and raw accounting variables in a single pipeline, enabling the model to derive new ratio-like features that capture real-world financial complexity.Comprehensive experiments demonstrate that this dual-layer cost weighting, together with SMOTE, significantly enhances fraud detection. Notably, incorporating TabTransformer within the ensemble achieves an AUC of 89.10% on a benchmark dataset, surpassing single-stage cost-sensitive methods and prior studies. These findings underscore how multi-level cost sensitivity, attention-driven modeling, and domain-specific feature integration can effectively tackle skewed data in financial statement fraud. Our framework offers an adaptable and robust solution for class-imbalanced financial contexts and may extend to related anomaly detection scenarios across various domains.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".