Boosting simple decision trees with Bayesian learning for text categorization

Introduces a Bayesian method to select best base classifiers for a boosting algorithm that is used for solving text categorization problems. This method is specifically shaped for an improved version of AdaBoost.MH, an effective multi-class multi-label text classification algorithm. The paper also proposes a method to facilitate its convergence. Experimental results show that these changes improve not only the accuracy, but also the efficiency of boosting algorithms for text categorization.