For listed companies, financial status is a very important factor for national macro-economy and local economic construction. Through Logistic regression and data survey and comparison, the analysis of accounting fraud of listed companies is studied. The analysis results are as follows: Comparing the three commonly used methods in the research method, Logistic regression analysis is the most used method at present, the accuracy is 20% higher than other commonly used methods, and the F1 performance index is more than 15% higher. After this method, the behavior analysis speed and analysis accuracy of accounting fraud have been greatly improved. Through the analysis, it can be seen that there are currently many means of accounting fraud, among which the most frequently used is false reporting of fraudulent amounts, and the degree of influence is also the largest, reaching an average of 90%. In order to reduce the impact, the corresponding preventive measures are put forward. After comparison, it is found that strengthening the legal system has the greatest impact and the best effect.