Workshop on Data Analysis

This paper presents a meta-learning approach for textual document classification task and an automatic selection of the best available algorithm for creation of classifiers. After brief introductory description of principles of creation and evaluation of the classifiers, the meta-learning approach is presented as a method for automatic selection of the most appropriate classifier algorithm for creation of binary classifiers. Designed methods, based on the modification of MUDOF (Meta-learning Using Document Feature Characteristics) algorithm, are described together with its implementation using the JBowl (Java Bag of word library). Finally, the experimental results achieved by the meta-learning algorithms as well as their comparisons with traditional ways used for text classification are presented.