Discretization of data using Boolean transformations and information theory based evaluation criteria
Year: 2015
Type: article
Abstract: Discretization is one of the most important parts of decision table preprocessing. Transforming continuous values of attributes into discrete intervals influences further analysis using data mining methods. In particular, the accuracy of generated predictions is highly dependent on the quality of discretization. The paper contains a description of three new heuristic algorithms for discretization of numeric data, based on Boolean reasoning. Additionally, an entropy-based evaluation of discretization is introduced to compare the results of the proposed algorithms with the results of leading university software for data analysis. Considering the discretization as a data compression method, the average compression ratio achieved for databases examined in the paper is 8.02 while maintaining the consistency of databases at 100%.
Authors: C. Jankowski, D. Reda, M. Mańkowski, G. Borowik
Institution: Warsaw University of Technology
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $283
OpenAlex ID: W2325382893