This study addresses the challenges of highly concealed abnormal behaviors, scarce fraud samples, and complex high-dimensional feature structures in credit card transactions by proposing a fraud detection framework that integrates a generative prior. The method first uses a generative model to learn the latent distribution of normal transactions and builds a stable behavioral baseline, where distributional constraints strengthen the structural representation of normal patterns. The latent variables from the generative model are then embedded jointly with discriminative features, allowing the model to capture both local attributes and global behavioral structures within a unified representation space and thereby improving the detection of weak anomalies and structural deviations. To evaluate performance comprehensively, the study conducts multidimensional experiments, including hyperparameter sensitivity, environmental disturbance sensitivity, and data disturbance sensitivity, to analyze how the generative prior affects model stability and discriminative capability under different external conditions. The results show that the framework outperforms traditional methods across Accuracy, Precision, Recall, and AUC, and maintains higher robustness and consistency in noisy, imbalanced, and dynamic transaction environments. The findings demonstrate that integrating generative distribution modeling with discriminative risk representation significantly enhances the model's sensitivity to abnormal behaviors and provides a more structured and reliable solution for risk identification in complex financial settings.

This study proposes a method for automatic audit report generation and compliance analysis based on the BERT model, aiming to solve the problems of low efficiency, high labor costs, and insufficient compliance detection accuracy in traditional audit work. By combining the powerful semantic understanding ability of the pre-trained language model, this study achieves high-quality generation of audit reports and uses classification tasks to accurately identify potential risk points in the text. The experiment uses multiple public datasets to evaluate the performance of BERT in text generation quality (BLEU Score) and compliance detection (Accuracy, F1-score). Compared with models such as GPT, T5, and LSTM, BERT performs best in all indicators. At the same time, through robustness testing, the performance of the model in the face of text noise or tampered data is analyzed. The results show that although noise has a certain impact on model performance, BERT's overall performance is still better than that of other comparison models. This study also explores the impact of different fine-tuning strategies on model performance and analyzes its adaptability and limitations in the audit field. The research findings demonstrate that the BERT-based methodology not only effectively enhances the quality of the automated generation of audit reports but also substantially improves the accuracy of compliance analysis. This advancement offers novel insights and serves as a guiding framework for the development of intelligent audit technology.

This paper addresses the complexity and high-risk nature of fraud detection in financial transaction data by proposing a discrimination method based on an improved Mamba architecture. The study first performs data cleaning, normalization, and embedding to reduce scale differences across multimodal features, and employs temporal encoding and positional embedding to strengthen sequence representation. A dynamic state updating and selective state space mechanism is then introduced to capture long-term dependencies and global contextual features, thereby enhancing the model's discrimination ability in complex transaction environments. Furthermore, context interaction and weighted aggregation modules are designed to integrate hidden states across time steps into a global representation, providing more robust inputs for final classification. Experiments, including baseline comparisons, sequence length sensitivity, and noise robustness tests, demonstrate that the proposed method outperforms multiple mainstream architectures in Accuracy, F1-score, AUC, and Precision, while showing greater stability in handling data imbalance and adversarial perturbations. Overall, the study confirms the effectiveness and applicability of the improved Mamba architecture for financial fraud detection and provides technical support for building secure and trustworthy risk control systems.

This paper proposes a deep discriminative model based on causal representation learning for the task of fraud detection in financial transactions. The method aims to enhance model stability and generalization in complex environments, such as distribution shifts and strategy changes, by modeling the underlying causal structures among transaction features. Specifically, the model first captures the generative process of transaction data using causal structural modeling. It introduces latent variables to represent causal dependencies among generative factors. Next, a causal encoder and reconstruction decoder are constructed within a variational autoencoder framework. This enables causally guided representation learning from the input data. Based on these representations, a classifier is designed to perform the fraud detection task. A domain discriminator is also introduced to enforce domain invariance in the latent space. This helps improve robustness under diverse fraud strategies. To validate the proposed method, experiments are conducted on a publicly available financial fraud dataset. The model is evaluated from multiple perspectives, including discrimination ability, stability, and cross-domain generalization. Results show that the proposed model outperforms representative existing methods across several metrics. It shows particularly strong performance in scenarios with noticeable strategy changes or significant feature perturbations. These findings further demonstrate the modeling advantage of causal mechanisms in fraud detection tasks. The overall framework enhances the extraction of key fraud-related features. It also provides theoretical and practical value for representation learning methods in high-risk environments.

This study proposes a multi-asset portfolio risk prediction model that integrates Long Short-Term Memory (LSTM) networks with Copula functions. The model is designed to capture both the temporal dynamics of financial asset returns and the nonlinear dependencies among assets. LSTM is used to model the marginal distributions of individual asset return series. Copula functions are employed to describe the joint distribution structure across multiple assets. This allows for accurate prediction of key portfolio risk metrics such as Value-at-Risk (VaR) and Conditional Value-at-Risk (CVaR). In the experimental design, several baseline models are constructed for performance comparison. Further analyses are conducted to assess the model's risk prediction ability under varying numbers of assets and to evaluate risk coverage across different confidence intervals. The experimental results show that the proposed LSTM-Copula model outperforms mainstream methods across multiple evaluation metrics. It demonstrates higher robustness and predictive accuracy, particularly in high-dimensional and sparse data settings. This approach offers a novel path for financial risk management by combining statistical modeling with deep learning. It provides strong empirical results and holds substantial practical value for applications in complex financial environments.

This paper proposes a stock price prediction model based on a combination method of recurrent neural networks (RNN) and attention mechanisms. The goal is to improve the forecasting accuracy of financial time series data. By applying the attention mechanism to the RNN architecture, the model can learn dynamic importance weights for different time steps. This allows better capture of dominant features in stock price volatility. The study utilizes real Apple Inc.'s stock prices between January 2024 and January 2025 to build models and validate the models. Normalizing and implementing train-test strategies were used to preprocess data in a bid to enhance stability within the models. Experiments utilize mean squared error (MSE), mean absolute error (MAE), and the coefficient of determination (R2) as evaluation metrics for the model. Comparative evaluation was performed against a set of mainstream popular models. Results show that the constructed RNN+Attention model outperforms traditional RNN, LSTM, MLP, and Transformer models both in prediction accuracy and fitting performance. In addition, visualizations demonstrate a strong correspondence between predicted values and actual values. The curve of training loss shows a clear downward trend, further supporting the effectiveness and convergence of the model. This method provides an efficient and realistic technical solution to stock price prediction using deep learning.