Editing training data for multi-label classification with the k-nearest neighbor rule

Sawsan Kanj (1) , Fahed Abdallah (1) , Thierry Denoeux (2, 1) , Kifah Tout (3)

1 Heudiasyc - Heuristique et Diagnostic des Systèmes Complexes [Compiègne]
2 Labex MS2T - Laboratoire d'Excellence "Maîtrise des Systèmes de Systèmes Technologiques"
3 Azm Center for Biotechnology Research

Multi-label classification allows instances to belong to several classes at once. It has received significant attention in machine learning and has found many real world applications in recent years, such as text categorization, automatic video annotation and functional genomics, resulting in the development of many multi-label classification methods. Based on labelled examples in the training dataset, a multi-labelled method extracts inherent information in order to output a function that predicts the labels of unlabelled data. Due to several problems, like errors in the input vectors or in their labels, this information may be wrong and might lead the multi-label algorithm to fail. In this paper, we propose a simple algorithm for overcoming these problems by editing the existing training dataset, and adapting this edited set with different multi-label classification methods. Evaluation on benchmark datasets demonstrates the usefulness and effectiveness of our approach.

Mots clés: prototype selection; edition; Classification; multi-label; k-nearest neighbors rule

Domaines: Intelligence artificielle [cs.AI]

Sawsan Kanj, Fahed Abdallah, Thierry Denoeux, Kifah Tout. Editing training data for multi-label classification with the k-nearest neighbor rule. Pattern Analysis and Applications, 2016, 19 (1), pp.145-161. ⟨10.1007/s10044-015-0452-8⟩. ⟨hal-01294269⟩