A Hybrid Fuzzy Maintained Classification Method Based on Dendritic Cells

Abstract

The dendritic cell algorithm (DCA) is a classification algorithm based on the behavior of natural dendritic cells (DCs). In literature, DCA has given good classification results. However, DCA was known to be sensitive to the order of the instance classes. To solve this limitation, a fuzzy DCA version was developed stating that the cause of such sensitivity is related to the DCA crisp classification task (hypothesis 1). In this paper, we hypothesize that there is a second possible cause of such DCA sensitivity which is related to the possible existence of noisy instances presented in the DCA signal data set (hypothesis 2). Thus, we aim, first of all, to test the trueness of the latter hypothesis, and second, we aim to develop an overall hybrid DCA taking both hypotheses into consideration. Based on hypothesis 1, our new DCA focuses on smoothing the crisp classification task using fuzzy set theory. Based on hypothesis 2, a data set cleaning technique is used to guarantee the quality of the DCA signal data set. Results show that our proposed hybrid fuzzy maintained algorithm succeeds in obtaining results of interest.