Research on Anomaly Detection and Risk Assessment Model Based on GA‐SA Algorithm in Financial Audit

With the rapid development of information technology, traditional financial auditing methods have been difficult to meet the requirements of modern enterprises for efficiency and accuracy of financial auditing. Therefore, this paper proposes a financial audit anomaly detection and risk assessment model based on the combination of genetic algorithm and simulated annealing algorithm. The model integrates Genetic Algorithm and Simulated Annealing to combine GA's global exploration capability with SA's local convergence strength, thus enhancing the precision and stability of anomaly detection and risk assessment. In the study, we conducted an experimental analysis on the financial data of a certain enterprise. Financial data from 2019 to 2022 were selected, including key financial indicators such as total assets, net profit, cash flow, and debt ratio, totaling 4 years, 10 financial indicators, and 2000 transaction data. On this basis, the GA‐SA algorithm is used to process the data, and the anomaly detection accuracy rate of the model output reaches 92.3%, which is 12.5% higher than that of the traditional method. In addition, the performance of the model in risk assessment is also very outstanding. The accurate prediction rate of enterprise risk level is 90.5%, which is 8.7% higher than the traditional risk assessment model. Through the optimization calculation of the model, the prediction lead time of potential financial risks of enterprises is shortened by about 3 months, and the early warning efficiency is improved by 15%. Considering the multi‐dimensionality and complexity of financial data, GA‐SA algorithm not only improves the efficiency of financial audit, but also effectively enhances the audit quality. The experimental results show that the performance of GA‐SA model in anomaly detection and risk assessment is significantly better than that of traditional methods, supported by comprehensive empirical evaluations including comparisons with baseline models and sensitivity analyses of key parameters, demonstrating strong practical application value.