Random Feature Selection using Random Subspace Logistic Regression

Feature selection becomes a prominent method in the big data era. The logistic regression model is a wrapper method that provides better classification or prediction accuracy but it is computationally expensive. In this study, we propose the random subspace logistic regression where features are randomly selected through bootstrap cycles. The random subspace regression method is applied to both standard and lasso logistic regression models. Using the simulated and empirical data, our proposed random subspace logistic regression shows favorable results and can be a promising alternative for flat feature selection