Deep reinforcement learning-based heterogeneous graph neural networks for multi-task financial fraud detection: a new framework incorporating variational self-encoders
Bibliographic record
Abstract
Financial fraud, as a global problem in the financial industry, brings huge economic losses to financial institutions and customers.In this paper, a multi-task financial fraud detection model is constructed based on heterogeneous graph neural network with deep reinforcement learning, combined with variational self-encoder.In this model, the variational self-encoder is combined with graph convolutional network to construct the node input representation coding module, as a way to enhance the multi-task financial fraud data and better mine the structured features of different nodes.The attention mechanism is then introduced to build the relation-aware attention, which deeply mines the input node features, further acquires the neighbor-generated features of different nodes in the network, and combines the mutual information to measure the nonlinear correlation between different random nodes.Then the financial fraud node representation is mapped into the highdimensional space by the multilayer perceptron, and then the financial fraud prediction confidence of the model is obtained, and different types of loss functions are set to ensure the detection efficiency of the model.The results show that the F1-macro and AUC values of the financial fraud detection model on the self-constructed FFD dataset are 0.749 and 0.925, respectively.Relying on the heterogeneous graphical neural network and the variational autocoder, a multi-task financial fraud detection model can be constructed, which provides a new idea for solving the suspected fraud and money laundering cases that may exist in the field of finance and economy.
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 | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".