Optimization of decision rule complexity for decision tables with many-valued decisions
Type
Conference Paper
KAUST Department
Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) DivisionApplied Mathematics and Computational Science Program
Extensions of Dynamic Programming, Machine Learning and Discrete Optimization Research Group
Date
2013-10Permanent link to this record
http://hdl.handle.net/10754/564808Metadata
Show full item recordAbstract
We describe new heuristics to construct decision rules for decision tables with many-valued decisions from the point of view of length and coverage which are enough good. We use statistical test to find leaders among the heuristics. After that, we compare our results with optimal result obtained by dynamic programming algorithms. The average percentage of relative difference between length (coverage) of constructed and optimal rules is at most 6.89% (15.89%, respectively) for leaders which seems to be a promising result. © 2013 IEEE.Conference/Event name
2013 IEEE International Conference on Systems, Man, and Cybernetics, SMC 2013ISBN
9780769551548Scopus Count
Related items
Showing items related by title, author, creator and subject.
-
Decision rules for decision tables with many-valued decisions(Springer Science + Business Media, 2011)In the paper, authors presents a greedy algorithm for construction of exact and partial decision rules for decision tables with many-valued decisions. Exact decision rules can be 'over-fitted', so instead of exact decision rules with many attributes, it is more appropriate to work with partial decision rules with smaller number of attributes. Based on results for set cover problem authors study bounds on accuracy of greedy algorithm for exact and partial decision rule construction, and complexity of the problem of minimization of decision rule length. © 2011 Springer-Verlag.
-
Minimization of Decision Tree Average Depth for Decision Tables with Many-valued Decisions(Elsevier BV, 2014-09-13)The paper is devoted to the analysis of greedy algorithms for the minimization of average depth of decision trees for decision tables such that each row is labeled with a set of decisions. The goal is to find one decision from the set of decisions. When we compare with the optimal result obtained from dynamic programming algorithm, we found some greedy algorithms produces results which are close to the optimal result for the minimization of average depth of decision trees.
-
Multi-stage optimization of decision and inhibitory trees for decision tables with many-valued decisions(Elsevier BV, 2017-06-16)We study problems of optimization of decision and inhibitory trees for decision tables with many-valued decisions. As cost functions, we consider depth, average depth, number of nodes, and number of terminal/nonterminal nodes in trees. Decision tables with many-valued decisions (multi-label decision tables) are often more accurate models for real-life data sets than usual decision tables with single-valued decisions. Inhibitory trees can sometimes capture more information from decision tables than decision trees. In this paper, we create dynamic programming algorithms for multi-stage optimization of trees relative to a sequence of cost functions. We apply these algorithms to prove the existence of totally optimal (simultaneously optimal relative to a number of cost functions) decision and inhibitory trees for some modified decision tables from the UCI Machine Learning Repository.
