Evolutionary Approaches for the Extraction of Classification Rules

An Overwiew Of Our Proposed Methods

In this paper, we are interested in the classification rules extraction problem and will construct a classification model using hybrid and parallel evolutionary methods. Indeed, evolutionary approaches (EAs) are well known for their great power to explore the search space. They present the ability of escaping from local optima due to their inherent global search capability. Their concurrent search enables them to promptly explore and identify new promising regions of the solution space. Though EAs are capable of global exploration and locating new regions of the solution space to identify potential candidates, however, when a potential region is identified, there is no further focus on the micro-exploitation aspect. Hence, local search are often used as a complement to EAs optimization that concentrate mainly on global exploration. For this, memetic algorithms, which incorporate local improvement search into EAs, are proposed (Ong & Keane, 2004). Experimental studies have shown that EA-Local Search (LS) hybrids or hybrid EAs are capable of more efficient search capabilities (Merz & Freisleben, 1999).

In a first step, our paper provides a memetic approach to extract a classification model which will be constituted by a set of rules presenting a high accuracy degree. Hence we will present all the modeling details of an extractor of classification rules based on a genetic algorithm (GA) firstly, then a Tabu Search (TS) to extract classification rules will be presented and finally a memetic approach GA-TS-based will be detailed to solve the same problem.