Feature Reduct of Decision Tables Based on Feature Matrix
赵卫东,戴伟辉
DOI: https://doi.org/10.3321/j.issn:1000-6788.2003.03.012
IF: 2.034
2003-01-01
Systems Engineering
Abstract:Feature reduct of decision tables is important for rough analysis. To consistent decision tables, the minimal reduct has been proved to be NP-hard. Many heuristic algorithms, therefore, have been given but most of them depend on the core of decision tables, which is not easy to get, especially for large-scale decision tables. In this way, there exist some problems such as efficiency and the solution completeness in them. Based on feature matrix put forward for reduct of decision tables, a new method to solve the difficulty is proposed in this paper. The method not only is independent of the core of decision tables, but provides a way-out for it. Also, decision tables containing null values, which are regardless and hard to deal with at present, are analyzed in depth in order to find useful information.