Year: 2025
Type: conference-paper
Abstract: This paper addresses the issue of identifying financial fraud in enterprises and proposes an identification model based on the XGBoost algorithm. Using the financial data from the 9th 'Teddy Cup' Data Analysis Skills Competition as the basis, we preprocess the data by handling missing values and outliers, and perform correlation analysis using Pearson's correlation coefficient to identify five key financial indicators closely related to total profit. A financial fraud identification model is constructed based on the XGBoost algorithm. To address the data imbalance issue, we apply the undersampling method using the one-side selection technique, perform oversampling using the ADASYN method, and optimize the model parameters through grid search. The experimental results show that the XGBoost model achieves an accuracy of 0.9997726, a precision of 0.9714285, and a recall of 1.0. Compared with logistic regression and SVM models, the results indicate that XGBoost's predictive performance is significantly superior. This study provides new methods and tools for financial fraud detection and has important implications for financial risk assessment and early warning.
Authors: Lexing Deng, Hu Can, Yongtang Guan, Wubin Zeng, Wanxue Zeng, Lihao Wu
Institution: Guangzhou City University of Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412789780