Wyniki wyszukiwania - BazTech

1

Experimental Study of Totally Optimal Decision Trees

Aldilaijan Abdulla, Azad Mohammad, Moshkov Mikhail

Fundamenta Informaticae

|

2019

|

Vol. 165, nr 3-4

245--261

EN

In this paper, we present results of experimental studies related to the existence of totally optimal decision trees (which are optimal relative to two or more cost functions simultaneously) for nine decision tables from the UCI Machine Learning Repository. Such trees can be useful when we consider decision trees as algorithms for problem solving or as a way for knowledge representation. For cost functions, we use depth, average depth, and number of nodes. We study not only exact but also approximate decision trees based on five uncertainty measures: entropy, Gini index, misclassification error, relative misclassification error, and number of unordered pairs of rows with different decisions. To investigate the existence of totally optimal trees, we use an extension of dynamic programming that allows us to make multi-stage optimization of decision trees relative to a sequence of cost functions. Experimental results show that totally optimal decision trees exist in many cases. The behavior of graphs that describe how the number of decision tables with totally optimal decision trees depends on their accuracy is mainly irregular. However, one can observe some trends, in particular, an upward trend when accuracy is decreasing.

2

An Uncertainty Measure Based on Lower and Upper Approximations for Generalized Rough set Models

Wang Zhaohao, Yue Huifang, Deng Jianping

Fundamenta Informaticae

|

2019

|

Vol. 166, nr 3

273--296

EN

Uncertainty measures are an important tool for analyzing data. There is the uncertainty of a rough set caused by its boundary region in rough set models. Thus the uncertainty measurement issue is also an important topic for rough set theory. Shannon entropy has been introduced into rough set theory. However, there are relatively few studies on the uncertainty measure in generalized rough set models. We know that the boundary region of a rough set is closely related to the upper and lower approximations in rough set models. In this paper, from the viewpoint of the upper and lower approximations, we propose new uncertainty measures, the upper rough entropy and the lower rough entropy, in generalized rough set models. Then we focus on the investigations of the upper rough entropy, and give the concepts of the upper joint entropy, the upper conditional entropy and the mutual information with respect to a general binary relation. Some important properties of these measures are obtained. The connections among these measures are given. Furthermore, comparing with the existing uncertainty measures, the upper rough entropy has high distinguishing degree. Theoretical analysis and experimental results show that the proposed entropy is better effective than some existing measures.

3

A Comparative Study of MGRSs and their Uncertainty Measures

Ma Z.-M., Mi J.-S.

Fundamenta Informaticae

|

2015

|

Vol. 142, nr 1-4

161--181

EN

Multi-granulation rough set(MGRS), as a kind of fusion mechanism of different information or data, is an useful development of Pawlak rough set theory. Firstly, this paper gives an introduction for various types of MGRS, their properties and axiomatization characterizations are studied. We show that, except for the optimistic one, each of the existing MGRS means a single granulation rough set. Then, we made a comparative analysis on the different uncertainty measures among the various multi-granulation approximation spaces. At the basis of investigating for the existing uncertainty measures, we discuss their limitations via some examples, and propose a total ordered relation among approximation spaces, even in the more general covering ones. It will be better than the original partial relation in revealing uncertainty, which conceal in the approximation space or covering one. Finally, based on the total ordered relation, we present improved information entropy, rough entropy, knowledge granulation and axiomatic definition of the knowledge granulation measures. It is proved that they are more reasonable than the original ones. Then, some novel uncertainty measures and improved fusion uncertainty measures about various granulations are also proposed. By employing these measures, granulation measures of various MGRSs are defined and studied.