Multiple parameter control for ant colony optimization applied to feature selection problem

摘要: The ant colony optimization algorithm (ACO) was initially developed to be a metaheuristic for combinatorial problem. In scores of experiments, it is confirmed that the parameter settings in ACO have direct effects on performance algorithm. However, few studies specially reported control ACO. aim this paper put forward some strategies adaptively adjust and further provide deeper understanding control, including static dynamic parameters. We choose well-known system (AS) (ACS) controlled by our proposed strategies. parameters AS ACS include β, pheromone evaporation rate (?), exploration probability factor (q0) number ants (m). three adaptive (SI, SII SIII) based fuzzy logic which adjusts ?, q0 m, respectively. feature selection problem considered evaluating addition, because are not intrinsically fit problem, we modified ACS, named as (FAAS) (FAACS), make them more suitable Because only one allowed dynamically adjusted FAAS or FAACS, remaining should statically specified. Thus, parametric guidelines proper combination settings. FAACS compared with AS-based, ACS-based, particle swarm optimization-based genetic algorithm-based methods comprehensive set 10 benchmark data sets, taken from UCI machine learning StatLog databases. numerical results statistical analysis show algorithms outperform significantly than other terms prediction accuracy smaller subset features.