Generating and Postprocessing of Biclusters from Discrete Value Matrices
- Cite this paper as:
- Michalak M., Stawarz M. (2011) Generating and Postprocessing of Biclusters from Discrete Value Matrices. In: Jędrzejowicz P., Nguyen N.T., Hoang K. (eds) Computational Collective Intelligence. Technologies and Applications. ICCCI 2011. Lecture Notes in Computer Science, vol 6922. Springer, Berlin, Heidelberg
This paper presents a new approach for the biclustering problem. For this purpose new notions like half-bicluster and biclustering matrix were developed. Results obtained with the algorithm BicDM (Biclustering of Discrete value Matrix) were compared with some other methods of biclustering. In this article the new algorithm is applied for binary data but there is no limitation to use it for other discrete type data sets. In this paper also two postprocessing steps are defined: generalization and filtering. In the first step biclusters are generalized and after that only those which are the best become the final set - weak biclusters are filtered from the set. The usage of the algorithm makes it possible to improve the description of data with the reduction of bicluster number without the loss of information. The postprocessing was performed on the new algorithm results and compared with other biclustering methods.
Keywordsmachine learning data mining biclustering postprocessing
Unable to display preview. Download preview PDF.