Fuzzy LASSO Logistic Regression: A Regularized Framework for Interpretable Binary Classification Under Uncertainty

Fuzzy classification models are important for handling uncertainty and heterogeneity in high-dimensional data. Although recent fuzzy logistic regression approaches have demonstrated promising results, challenges remain in managing multicollinearity and performing effective feature selection for mixed-type predictors. This study proposes a Fuzzy LASSO Logistic Regression (FLLR) framework that integrates triangular fuzzy membership functions with LASSO regularization to address these issues. The proposed approach fuzzifies continuous, ordinal, and categorical features, performs variable selection via LASSO logistic regression, and mitigates multicollinearity through Variance Inflation Factor (VIF) screening. The model was trained using 10-fold cross-validation, and its performance was evaluated using four key metrics: Accuracy, F1-score, Matthews Correlation Coefficient (MCC), and Area Under the ROC Curve (AUC). Experiments on multiple benchmark datasets demonstrated that FLLR achieves strong predictive performance while preserving model interpretability. Comparative analyses with existing fuzzy logistic regression models and a standard artificial neural network further confirmed the robustness and practical utility of the proposed framework for uncertainty-aware decision-making using mixed-type data.