Seeking better solutions for clustering? This paper introduces a novel approach called the Constructive Genetic Algorithm (CGA) to solve optimization problems, particularly in the context of clustering. The CGA evaluates schemata and structures on a common basis through a bi-objective optimization process, enhancing the genetic algorithm's (GA) behavior. Problems are modeled to consider the evaluation of two fitness functions (fg-fitness). Evolution considers an adaptive rejection threshold that contemplates both objectives and ranks each individual in the population. The population is dynamic in size and composed of schemata and structures. Recombination preserves good schemata, while mutation diversifies structures. The CGA is applied to two graph clustering problems: the classical p-median and the capacitated p-median. The results show that this new approach provides good solutions for instances taken from the literature, offering a new tool for tackling complex clustering challenges.
Published in _Evolutionary Computation_, this paper is highly relevant to the journal's focus on evolutionary algorithms and computational optimization techniques. The introduction of the Constructive Genetic Algorithm (CGA) for clustering problems aligns with the journal's scope. The study demonstrates the effectiveness of the CGA by applying it to classical clustering problems, showcasing its potential in the field of evolutionary computation.