Construction and Simulation of Financial Audit Model Based on Convolutional Neural Network.

Area of Science:

• Computer Science
• Data Science
• Auditing

Background:

• Big data revolutionizes information processing, making data analysis crucial for comprehensive audit and supervision.
• Convolutional Neural Networks (CNNs) show promise for financial auditing due to their feature extraction capabilities.
• CNNs face challenges like gradient disappearance and convergence issues, limiting their effectiveness in financial audits.

Purpose of the Study:

• To improve the performance of CNN-based financial audit models.
• To address the limitations of standard CNNs in financial audit applications.
• To explore the integration of genetic algorithms for CNN optimization in auditing.

Main Methods:

• Applied genetic algorithms to optimize the initial weights of convolutional neural networks.
• Investigated error sensitivity and learning rate impacts on different hidden layers.
• Analyzed the influence of learning rates on CNN convergence speed.
• Compared recognition performance against other algorithms on financial audit datasets.

Main Results:

• The optimized CNN model demonstrated superior performance in financial audit tasks.
• The improved learning rate algorithm resulted in a lower recognition error rate compared to standard CNNs.
• The genetic algorithm optimization effectively addressed CNN convergence and gradient issues.

Conclusions:

• Genetic algorithm-optimized CNNs offer enhanced performance for financial auditing.
• The proposed method improves the accuracy and efficiency of big data analysis in audits.
• This approach represents a significant advancement in intelligent auditing techniques.