Abstract
The detection of corporate accounting fraud is a critical challenge in the financial industry, where traditional models such as neural networks, logistic regression, and support vector machines often fall short in achieving high accuracy due to the complex and evolving nature of fraudulent activities. This paper proposes an enhanced approach to fraud detection by integrating convolutional neural networks (CNN) and long short-term memory (LSTM) networks, complemented by an attention mechanism to prioritize relevant features. To further improve the model's performance, the sparrow search algorithm (SSA) is employed for parameter optimization, ensuring the best configuration of the CNN-LSTM-Attention framework. Experimental results demonstrate that the proposed model outperforms conventional methods across various evaluation metrics, offering superior accuracy and robustness in recognizing fraudulent patterns in corporate accounting data.