Adaptive Affinity Functions by Ellipsoidal Artificial Lymphocytes in an Artificial Immune System for Classification

K. Igawa and H. Ohashi (Japan)


Artificial immune systems, artificial negative selection classifier, classification, affinity function, ellipsoid


Artificial immune systems (AIS) are a type of intelligent system inspired by the mechanisms and principles observed in the human immune system. Recently, methods for designing an effective AIS for data mining have been discussed. The outcome of this discussion is that each component of the AIS should be tailored to the application. The affinity function is an important part of an AIS because it defines the degree of interaction between artificial cells and data. However, choosing an appropriate affinity function is a difficult task because it necessitates the understanding of the data set in detail. In this paper, we propose an adaptation mechanism for the affinity function to be used in classification. For this purpose, we introduce ellipsoidal artificial lymphocytes (eALCs) into an Artificial Negative Selection Classifier (ANSC). Each eALC has its own affinity function that can be adapted automatically to each problem. Experiments are conducted with the ANSC in conjunction with eALCs using artificial and real-world data sets. The results show that eALCs and the adaptation mechanism are useful for classification.

Important Links:

Go Back