A genetic algorithm-based dendritic cell algorithm for input signal generation

The dendritic cell algorithm (DCA) is a classification algorithm based on the biological antigen presentation process. Its classification efficiently depends on a data preprocessing procedure, where feature selection and signal categorization are the main work for generating input signals. Several methods have been employed (e.g., correlation coefficient and rough set theory). Those studies preferred to measure the importance of features by evaluating their relevance to the class. Generally, they determined a mapping relationship between important features and signal categories of DCA based on expert knowledge. Typically, those studies ignore the effect of unimportant features, and the mapping relationship determined by expertise may not produce an optimal classification result. Thus, a hybrid model, GA-DCA, is proposed for feature selection and signal categorization based on the genetic algorithm (GA). This study transforms feature selection and signal categorization into a grouping task (i.e., divides features into different signal groups). This study introduces a permutation-based expression with “Group" symbols to represent a potential feature grouping scheme. Correspondingly, adaptive operators are proposed to expand each possible scheme on the path from the initial feature grouping to the best feature grouping. GA-DCA searches the optimal feature subset and automatically assigns them to the most suitable signal groups without expertise. This study verifies the proposed approach by employing the UCI Machine Learning Repository and Keel-dataset Repository, and significant performance improvement is achieved.