A hybrid artificial immune classifier based on weighting attributes and fuzzy clustering

Inspired by complementary strategies, a hybrid supervised artificial immune classifier is put forward, which is on the basis of the clonal selection principle, and combined with the fuzzy c-means clustering (FCM) algorithm and information entropy theory. The new approach uses a weighted Euclidean distance based dissimilarity measure during all affinity evaluations. With the help of FCM clustering, the initial antibodies that image features of data set are extracted effectively, and then a clonal selection algorithm named CLONALG is adopted for each training instance to constitute the memory cells. Finally, classification is performed in a k-nearest neighbor approach with the developed set of memory cells. Experimental results on five benchmark datasets from UCI machine learning repository demonstrate the effectiveness of the algorithm as a classification technique. Compared with generic CLOALG artificial immune classifiers, the hybrid classifier not only can decrease the computational time, but also can achieve higher classification accuracy.