Financial Statement Fraud Detection based on Integrated Feature Selection and Imbalance Learning

Abstract

Based on the data mining technology, this paper proposes an integrated feature selection method to construct the financial statement fraud detection feature system of listed companies, uses the SMOTE algorithm to solve the class unbalanced distribution problem, and combines the machine learning algorithm models to construct the financial statement fraud detection model. Based on the real financial statement data of Chinese listed companies, the empirical analysis is conducted to provide support for the auditors. The integrated feature selection framework proposed in this paper improves the problem of poor generalization of the single feature selection method, and the SMOTE effectively strengthens and improves the ability of the model to detect financial statement fraud of listed companies.