Using real-world transaction data to identify money laundering: Leveraging traditional regression and machine learning techniques
Bibliographic record
Abstract
Money laundering is a pervasive legal and economic problem that hides criminal activity. Identifying money laundering is a priority for both banks and governments, thus, machine learning algorithms have emerged as a possible strategy to detect suspicious financial activity within financial institutions. We used traditional regression and supervised machine learning techniques to identify bank customers at an increased risk of committing money laundering. Specifically, we assessed whether model performance differed across varying operationalizations of the outcome (e.g., multinomial vs. binary classification) and determined whether the inclusion of investigator-derived novel features (e.g., averages across existing features) could improve model performance. We received two proprietary datasets from Scotiabank, a large bank headquartered in Canada. The datasets included customer account information (N = 4,469) and customers’ monthly transaction histories (N = 2,827) from April 15, 2019 to April 15, 2020. We implemented traditional logistic regression, logistic regression with LASSO regularization (LASSO), K-nearest neighbours (KNN), and extreme gradient boosted models (XGBoost). Results indicated that traditional logistic regression with a binary outcome, conducted with investigator-derived novel features, performed the best with an F1 score of 0.79 and accuracy of 0.72. Models with a binary outcome had higher accuracy than the multinomial models, but the F1 scores yielded mixed results. For KNN and XGBoost, we observed little change or worsening performance after the introduction of the investigator-derived novel features. However, the investigator-derived novel features improved model performance for LASSO and traditional logistic regression. Our findings demonstrate that investigators should consider different operationalizations of the outcome, where possible, and include novel features derived from existing features to potentially improve the detection of customers at risk of committing money laundering.
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 | 0.005 | 0.016 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.005 | 0.004 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| 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".