A niching genetic programming-based multi-objective algorithm for hybrid data classification

摘要: This paper introduces a multi-objective algorithm based on genetic programming to extract classification rules in databases composed of hybrid data, i.e., regular (e.g. numerical, logical, and textual) non-regular geographical) attributes. employs niche technique combined with population archive order identify the that are more suitable for classifying items amongst classes given data set. The is implemented such way user can choose function set adequate application. feature makes proposed approach virtually applicable any kind problem. Besides, problem modeled as one, which maximization accuracy minimization classifier complexity considered objective functions. A different problems, considerably sets domains, has been considered: wines, patients hepatitis, incipient faults power transformers level development cities. In this last set, some attributes geographical, they expressed points, lines or polygons. effectiveness compared three other methods, widely employed classification: Decision Tree (C4.5), Support Vector Machine (SVM) Radial Basis Function (RBF). Statistical comparisons have conducted employing one-way ANOVA Tukey's tests, provide reliable comparison methods. results show achieved better all tested instances, what suggests it considerable range applications.