Analysis of Deep Learning Techniques to Facilitate Automation of Financial Statements for Accounting Programs in Higher Education Institutions

Abstract

Abstract The rapid development of information technology and mobile Internet technology has brought about innovations in the field of education. The application of financial statement automation analysis methods in the teaching of accounting majors in higher vocational colleges and universities is conducive to improving the quality of the training of accounting talents in higher vocational colleges and universities so as to accelerate the construction of the modern accounting industry college. This study proposes the application of relevant deep learning techniques in the process of automated analysis of financial statements in terms of recognition and analysis, such as text detection based on CTPN and text recognition technology based on CRNN-Attention, which makes the processing of financial statements more automated and intelligent. It is also proposed to utilize the K-means algorithm to perform cluster analysis on the model output to extract the financial status of the company. To evaluate the model’s performance, several financial statement images of 12 companies are sampled for example analysis. The experimental data shows that the model has a high percentage of image skew correction, e.g., the ratio of image skew correction for the income statement and cash flow statement is 100%. The recognition speed of the model is also faster, with the consumption time of the profit and loss statement recognition task being around 0.3 seconds in 20 tests. In the financial statement analysis session, each of the three clusters of companies under the K-means algorithm clustering has its characteristics, with the first cluster of companies having a stronger quality of earnings, the third cluster of companies having outstanding profitability, and the second cluster of companies having average indicator data.