Lattice-Based Classification With Mixed Numeric And Nominal Data

Wei Hu,Huanye Sheng
DOI: https://doi.org/10.1007/978-3-540-37256-1_85
2006-01-01
Intelligent Control and Automation
Abstract:Traditional classification methods are often designed for certain types of data. They cannot be directly applied to dataset with mixed numeric and nominal data. Only after numeric data was discretized or nominal data was encoded, could algorithms work. As data should accommodate to algorithm, such learning scheme is approach oriented to some extend. This paper presents a new data mining scheme called lattice-based learning (LBL), whose central idea is formulating algorithms using basic operations on lattice structure. Since both numeric and nominal data can be easily embedded into lattices, LBL algorithms are applicable to any dataset with mixed data. We detail lattice-based classification (LBC) algorithm in this paper. The performance of LBC has been studied on different datasets. Results show that LBC is an effective method for classification with mixed data and LBL learning is a promising scheme for data mining.
What problem does this paper attempt to address?