With the development of the Internet, people are more and more pursuing the high-quality development of society, and the cultivation of high-quality talents is also more and more important. Aiming at the dilemma of professional ethics of senior accounting professionals, this paper introduces generator and discriminator structure based on a deep clustering algorithm and constructs a double autocoding deep embedded clustering model, i.e. DADEC model. After completing the model construction, iterative optimization of the clustering center is carried out by combining the small batch stochastic gradient descent and backpropagation algorithms to reduce the clustering loss. The optimized model is utilized in the professional ethics education of senior accounting professionals to categorize their dilemmas and explore breakthrough strategies. The results show that the perceptions of accounting majors in 15 higher vocational colleges and universities in terms of ethical knowledge, sense of professional dedication, professional ethical will, and professional ethical conduct are below 0.6 points. The mean value of the classroom effect of professional ethics education is 0.588, and there is a dilemma between a single teaching form and outdated teaching content. The scores of students in honesty, trustworthiness, love, and dedication are below 60 points. Although there are certain challenges in the current training of senior accounting professionals, the combination of this study points out the direction for overcoming these challenges.