Research on Constructing a Logistic-Based Enterprise Financial Risk Early Warning Model in the Context of Big Data

Addressing the complex dimensions and imbalanced distribution of enterprise financial data in risk identification, this study systematically investigates the construction of logistic regression classification models and feature engineering methods. It outlines the data cleansing, indicator selection, feature derivation, model training, and parameter tuning processes, establishing a structured modeling framework for financial risk early warning. By incorporating L1 regularization feature selection and SMOTE sample balancing strategy, a stable set of input variables was formed, and parameter optimization was achieved through cross-validation. Comparative experiments demonstrate that the logistic regression model achieves an Accuracy of 0.861, an F1-score of 0.804, and an AUC of 0.871, outperforming Support Vector Machine and Random Forest models in overall performance. Under SMOTE processing, Recall improves to 0.824, significantly enhancing risk sample identification capability. The findings validate the critical role of feature reduction and data balancing mechanisms in constructing classification boundaries.