Construction of Financial Fraud Risk Assessment Model Assisted by Artificial Intelligence

As the global economy rapidly expands, financial fraud has become increasingly common, causing substantial losses to investors and businesses. Traditional risk assessment methods, reliant on manual analysis, struggle to handle the growing complexity and volume of financial data. This study leverages financial data from publicly listed companies and employs artificial intelligence (AI) to create a more objective and robust financial fraud risk assessment model. Initially, financial data are gathered, cleaned, and standardized. Key indicators of fraud risk are then identified, including financial metrics, corporate governance, market performance, and non-financial factors. A novel model based on the random forest (RF) algorithm is developed, incorporating a dynamic weighting strategy to enhance predictive accuracy and generalization. Experimental results reveal that the proposed dynamic weighted RF model surpasses traditional RF, logistic regression (LR), and support vector machine (SVM) models in accuracy, recall, and F1 score. This model effectively detects fraud risks and is adaptable and interpretable across various types and sizes of company data.