Paper: arXiv 2502.15822
Authors: Tianzuo Hu
Abstract
This paper proposes a financial fraud detection system based on improved Random Forest (RF) and Gradient Boosting Machine (GBM). Specifically, the system introduces a novel model architecture called GBM-SSRF (Gradient Boosting Machine with Simplified and Strengthened Random Forest), which cleverly combines the powerful optimization capabilities of the gradient boosting machine (GBM) with improved randomization. The computational efficiency and feature extraction capabilities of the Simplified and Strengthened Random Forest (SSRF) forest significantly improve the performance of financial fraud detection. Although the traditional random forest model has good classification capabilities, it has high computational complexity when faced with large-scale data and has certain limitations in feature selection. As a commonly used ensemble learning method, the GBM model has significant advantages in optimizing performance and handling nonlinear problems. However, GBM takes a long time to train and is prone to overfitting problems when data samples are unbalanced. In response to these limitations, this paper optimizes the random forest based on the structure, reducing the computational complexity and improving the feature selection ability through the structural simplification and enhancement of the random forest. In addition, the optimized random forest is embedded into the GBM framework, and the model can maintain efficiency and stability with the help of GBM’s gradient optimization capability. Experiments show that the GBM-SSRF model not only has good performance, but also has good robustness and generalization capabilities, providing an efficient and reliable solution for financial fraud detection.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 6.5/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper introduces algorithmic modifications and ensemble methods, which involve some mathematical notation (e.g., GBM gradient formulas), but the depth is limited to applied machine learning concepts without advanced theoretical derivations. Empirically, it focuses on fraud detection with claims of performance improvements, robustness, and generalization, suggesting backtest-ready implementation with data/feature engineering, though specific dataset details or code are not provided in the excerpt.
Research Flowchart
flowchart TD
A["Research Goal:<br>Improve Financial Fraud Detection"] --> B{"Data & Input:<br>General Financial / Fraud Data"}
B --> C["Methodology:<br>Proposed GBM-SSRF Model"]
C --> D["Component 1:<br>Simplified & Strengthened RF<br>Reduced Complexity & Enhanced Feature Selection"]
C --> E["Component 2:<br>Gradient Boosting Machine<br>Optimization & Handling Nonlinear Problems"]
D & E --> F["Computational Process:<br>Embedded GBM-SSRF Architecture<br>Optimization via Gradient Boosting"]
F --> G{"Evaluation"}
G --> H["Key Findings / Outcomes:<br>High Accuracy<br>Robust Generalization<br>Computational Efficiency"]
style A fill:#f9f,stroke:#333,stroke-width:2px
style H fill:#9f9,stroke:#333,stroke-width:2px
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a financial fraud detection system called GBM-SSRF, which combines an improved Random Forest (SSRF) with Gradient Boosting Machine.
Source sentence
Specifically, the system introduces a novel model architecture called GBM-SSRF (Gradient Boosting Machine with Simplified and Strengthened Random Forest), which cleverly combines the powerful optimization capabilities of the gradient boosting machine (GBM) with improved randomization.
Assumes 1
- Normal transactions dominate the data and fraudulent transactions are rare (about 0.2%). data
Source sentence
Normal transactions account for the vast majority, and the proportion of fraudulent transactions is low (usually about 0.2%).
Data
- credit card fraud detection datasets alternative public
Source sentence
The data sources include credit card fraud detection datasets on the Kaggle platform and desensitized actual bank transaction data.
- desensitized actual bank transaction data alternative unknown
Source sentence
The data sources include credit card fraud detection datasets on the Kaggle platform and desensitized actual bank transaction data.
Costs
Not applicable
Finds 5
- GBM-SSRF achieves an accuracy of 99.72%, higher than RF (97.85%), XGBoost (98.52%) and traditional GBM (98.71%). performance
Source sentence
the accuracy of the GBM-SSRF model reaches 99.72%, which is significantly better than the 97.85% of Random Forest (RF), 98.52% of XGBoost and 98.71% of traditional GBM
- GBM-SSRF attains a precision of 94.83%, surpassing RF (89.12%), XGBoost (91.65%) and GBM (92.34%). predictive-accuracy
Source sentence
In terms of precision, the accuracy of the GBM-SSRF model is as high as 94.83%. In comparison, the accuracy of RF, XGBoost and traditional GBM are 89.12%, 91.65% and 92.34% respectively.
- GBM-SSRF records a recall of 93.45%, higher than RF (87.34%), XGBoost (90.43%) and GBM (91.89%). predictive-accuracy
Source sentence
At the same time, GBM-SSRF also performs well in recall, with a recall rate of 93.45%, which is significantly higher than RF's 87.34%, XGBoost's 90.43% and traditional GBM's 91.89%.
- GBM-SSRF achieves an F1‑Score of 94.14%, outperforming RF (88.22%), XGBoost (91.04%) and GBM (92.11%). predictive-accuracy
Source sentence
The F1 score of the GBM-SSRF model reaches 94.14%, which is better than RF's 88.22%, XGBoost's 91.04% and traditional GBM's 92.11%.
- GBM-SSRF obtains an AUC‑ROC of 0.987, exceeding RF (0.942), XGBoost (0.956) and GBM (0.961). performance
Source sentence
Finally, in the AUC-ROC curve evaluation, the AUC value of GBM-SSRF was 0.987, which was significantly higher than 0.942 of RF, 0.956 of XGBoost and 0.961 of traditional GBM.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 90 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.