Research on internal financial fraud identification model of enterprise based on ensemble learning

In recent years, financial fraud cases have been on the rise, prompting numerous scholars to explore relevant fields and contribute significantly to the practical oversight of the economy. Integrated learning models have also gained widespread application in the realm of financial fraud detection, proving their efficacy in identification. This paper provides a summary of existing research and methodologies employed by scholars. After reviewing pertinent literature, the Logistic Regression model, a Single Decision Tree, Gradient Boosting Decision Trees, Random Forest model, XGBoost model, and LightGBM model were selected as candidate models for studying financial fraud detection. A comparative analysis of their respective identification accuracies was conducted. The research findings indicate that across the overall detection models, the identification rates of all models exceed 70%. Among these, the XGBoost model exhibits the best performance, achieving an identification accuracy of 87.77%. From the comparative results, it is evident that the accuracy of ensemble learning models generally surpasses that of traditional classification models and basic machine learning models, effectively enhancing the efficiency of financial fraud detection. Furthermore, in terms of identification speed, ensemble learning models demonstrate advantages such as shorter processing times and the ability to accommodate larger datasets.