Dynamic programming approach to optimization of approximate decision rules

A Round Trip Pattern for Building Decision Trees

“…Different strategies are employed for decision tree generating of different kinds of decision tables, e.g., single-valued decision tables [8][9][10][11] and multi-valued tables [12][13]. Construction algorithms roughly match some design pattern like dynamic programming [13], incremental algorithm [10] and greedy algorithm [11][12], etc. Normally, proposed algorithms attempt to construct decision trees of minimal height and minimal leaf number which are referred as time and spacial complexities.…”

Relationships Between Length and Coverage of Decision Rules

“…There are many different approaches to the design and analysis of decision rules: brute-force approach, genetic algorithms [25], Boolean reasoning [18,20,24], ant colony optimization [14], algorithms based on decision tree construction [15,17,21], algorithms based on a sequential covering procedure [7,10,11], different kinds of greedy algorithms [16,18], and dynamic programming [2,3,5,6,26]. We can find many programs which allow data analysis using decision rules created based on the approaches mentioned above, e.g., LERS [12], RSES [9], Rosetta [19], Weka [13], TRS library [23], and others.…”

Classifiers Based on Optimal Decision Rules

“…There are two aims for decision rule induction: construction of rule based classifiers and construction of rules that are good from the point of view of knowledge discovery and representation. There are different approaches to the construction of rules since the objectives can be, for example, Boolean reasoning [14,15,16], algorithms based on a sequential covering procedure [7,9,10], genetic algorithms [5,17], ant colony optimization [8,11], different kinds of greedy algorithms [12,14], dynamic programming approach [2,3,4,13]. The most part of mentioned approaches (with the exception of Boolean reasoning and dynamic programming) cannot guarantee the construction of shortest rules or rules with the maximum coverage.…”

“…In this paper, authors present results of the study of classifiers based on optimal exact and approximate decision rules constructed by dynamic programming algorithms [3,4]. To work with approximate decision rules, we use an uncertainty measure R(T ) that is the number of unordered pairs of rows with different decisions in the decision table T .…”

“…

Based on dynamic programming approach we design algorithms for sequential optimization of exact and approximate decision rules relative to the length and coverage [3,4]. In this paper, we use optimal rules to construct classifiers, and study two questions: (i) which rules are better from the point of view of classification -exact or approximate; and (ii) which order of optimization gives better results of classifier work: length, length+coverage, coverage, or coverage+length.

…”