Multi-label classification with Kohonen Maps and Winning Neighborhoods

• J. G. M. Barbirato UFSCar
• R. Cerri UFSCar

DOI: https://doi.org/10.5753/kdmile.2018.27398

Abstract

The conventional problem of classification in the context of machine learning is to classify examples of data sets in predefined categories, according to one or more similar characteristics. However, some data sets have classes with intersections, that is, examples can belong to more than one class simultaneously. Examples of these problems can be found, for example, in the identification of book genres and in the classification of images. These types of problems are called multi-label. The purpose of this article is to propose a new method of multi-label classification with Kohonen Maps. The idea is to use the winning neuron in the competitive process of the self-organizing map, together with the neighborhood around that neuron, for data classification. Thus, a new example is classified in the classes belonging to the training examples mapped for the winning neuron and its neighborhood. The Python language and the Scikit-Learn machine learning library were used to implement the neural network model, to implement evaluation measures, and to generate synthetic data sets. The use of a neighborhood of neurons was compared with a previous proposal using only one winning neuron. The results showed that the use of a neighborhood around the winning neuron is promising, obtaining better results.

Keywords: machine learning, classification, kohonen maps, multilabel, neural networks, self-organizing maps

References

Borges, H. and Nievola, J. Multi-label hierarchical classification using a competitive neural network for protein function prediction. In International Joint Conference on Neural Networks. pp. 1–8, 2012.

Cerri, R., Barros, R. C., P. L. F. de Carvalho, A. C., and Jin, Y. Reduction strategies for hierarchical multi-label classification in protein function prediction. BMC Bioinformatics 17 (1): 373, Sep, 2016.

Colombini, G. G., de Abreu, I. B. M., and Cerri, R. A self-organizing map-based method for multi-label classification. In Neural Networks (IJCNN), 2017 International Joint Conference on. IEEE, pp. 4291–4298, 2017.

Godbole, S. and Sarawagi, S. Discriminative methods for multi-labeled classification. In Pacific-Asia conference on knowledge discovery and data mining. Springer, pp. 22–30, 2004.

Mao, J., Huang, J., Toshev, A., Camburu, O., Yuille, A. L., and Murphy, K. Generation and comprehension of unambiguous object descriptions. In Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 11–20, 2016.

Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research vol. 12, pp. 2825–2830, 2011.

Tsoumakas, G., Katakis, I., and Vlahavas, I. Mining multi-label data. In Data mining and knowledge discovery handbook. Springer, pp. 667–685, 2009.

Wehrmann, J., Barros, R. C., Dôres, S. N. d., and Cerri, R. Hierarchical multi-label classification with chained neural networks. In Proceedings of the Symposium on Applied Computing. SAC ’17. ACM, New York, NY, USA, pp. 790–795, 2017.

Wehrmann, J., Cerri, R., and Barros, R. Hierarchical multi-label classification networks. In Proceedings of the 35th International Conference on Machine Learning, J. Dy and A. Krause (Eds.). Proceedings of Machine Learning Research, vol. 80. PMLR, Stockholmsmässan, Stockholm Sweden, pp. 5225–5234, 2018.

Zhang, M.-L. and Zhou, Z.-H. Multilabel neural networks with applications to functional genomics and text categorization. IEEE transactions on Knowledge and Data Engineering 18 (10): 1338–1351, 2006.