A hybrid model of fuzzy ARTMAP and genetic algorithm for data classification and rule extraction

摘要: A hybrid model (QFAM-GA) for data classification and rule extraction is proposed.Fuzzy ARTMAP (FAM) with Q-learning first used incremental learning of data.A Genetic Algorithm (GA) then feature selection extraction.Pruning to reduce the network complexity facilitate extraction.The results show QFAM-GA can provide useful if-then explain its predictions. two-stage proposed. The stage uses a Fuzzy classifier (known as QFAM) samples, while second from QFAM. Given new sample, resulting model, known QFAM-GA, able prediction pertaining target class sample well give fuzzy prediction. To complexity, pruning scheme using Q-values applied number prototypes generated by 'don't care' technique employed minimize input features GA. benchmark problems are evaluate effectiveness in terms test accuracy, noise tolerance, (number rules total length). comparable, if not better, than many other models reported literature. main significance this research usable intelligent (i.e., QFAM-GA) noisy conditions capability yielding set explanatory minimum antecedents. In addition, maximize accuracy simultaneously. empirical outcome positively demonstrate potential impact practical environment, i.e., providing an accurate concise justification domain users, therefore allowing users adopt decision support tool assisting their decision-making processes.