Optimized Neural Network Algorithm for Automatic Extraction of Financial Indicators and Its Application

The conventional method of manually extracting financial indicators is no longer sufficient to meet the needs of enterprises for efficient financial management. An efficient model for extracting financial indicators is proposed based on a multi-layer perceptron (MLP), with improvements achieved through optimized network architecture, tailored activation functions (ReLU and linear), and He initialization. The experimental results indicate that the model achieves a 92.2% accuracy rate on the test dataset, outperforming conventional approaches by roughly 15%, with the processing duration cut down to just one-fifth of the conventional methods. This study delves into the potential applications of neural network optimization algorithms within finance, accounting, and auditing sectors. In finance, it facilitates real-time data analysis, aiding in risk assessment and investment choices. In accounting, it enhances the efficiency of financial statement reviews and budget planning. As for auditing, it supports analytical processes, ensuring reliability and adherence to standards of financial information. With the development of technology, more intelligent techniques will be integrated into the proposed model, which will provide intelligent decision support for enterprise managers and promote the standardization of financial data analysis technology.