Hierarchical Text Classifier Using Ensemble Learning

The problem of data organization and classification on the Internet is one of the most actual topics in data science nowadays. This area of research is vividly developing due to growing amount of information on the Internet and high demand for its automatic processing. Therefore, a lot has been done in this sphere during the last years, but many problems are still to be solved. A good example of its subtask is the following: hierarchical data classification. Solution to the problem in question can be useful in tackling wide range of issues from creating a knowledge base to marketing research. The aim of this work is to develop a hierarchical website classification program based on ensemble learning. This paper provides an overview of existing methods for hierarchical website classification and proposes new algorithms and techniques to improve them. The paper proposes the approach of using weighted markup tags in feature extraction process. As a classifier we use an ensemble of classifiers combining two different hierarchical classification approaches. Performance of proposed algorithms is tested and compared with existing methods through several experiments. The Yandex Catalogue is used as a dataset for learning and testing the system. Keywords: hierarchical website classification, web mining, ensemble learning, feature extraction