A Novel Approach for Identifying Banded Patterns in Zero-One Data Using Column and Row Banding Scores

Abdullahi, Fatimah Binta, Coenen, Frans ORCID: 0000-0003-1026-6649 and Martin, Russell ORCID: 0000-0002-7043-503X
(2014) A Novel Approach for Identifying Banded Patterns in Zero-One Data Using Column and Row Banding Scores. In: Machine Learning and Data Mining in Pattern Recognition, 2014-7-21 - 2014-7-24, St. Petersburg, Russia.

This is the latest version of this item.

[img] Text
mldm2014abdullahi.pdf - Author Accepted Manuscript

Download (3MB)


Zero-one data is frequently encountered in the field of data mining. A banded pattern in zero-one data is one where the attributes (columns) and records (rows) are organized in such a way that the "ones" are arranged along the leading diagonal. The significance is that rearranging zero-one data so as to feature bandedness enhances the operation of some data mining algorithms that work with zero-one data. The fact that a dataset features banding may also be of interest in its own right with respect to various application domains. In this paper an effective banding algorithm is presented designed to reveal banding in 2D data by rearranging the ordering of columns and rows. The challenge is the large number of potential row and column permutations. To address this issue a column and row scoring mechanism is proposed that allows columns and rows to be ordered so as to reveal bandedness without the need to consider large numbers of permutations. This mechanism has been incorporated into the Banded Pattern Mining (BPM) algorithm proposed in this paper. The operation of BPM is fully discussed. A Complete evaluation of the BPM algorithm is also presented clearly indicating the advantages offered by BPM with respect to a number of competitor algorithms in the context of a collection of UCI Datasets. © 2014 Springer International Publishing Switzerland.

Item Type: Conference or Workshop Item (Unspecified)
Uncontrolled Keywords: Banded Patterns, Zero-One data, Data Mining and Pattern Mining
Depositing User: Symplectic Admin
Date Deposited: 13 Feb 2017 10:35
Last Modified: 07 Mar 2023 03:54
DOI: 10.1007/978-3-319-08979-9_5
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3005754

Available Versions of this Item