Artificial Immune Network Clustering Based on a Cultural Algorithm

Abstract

There are some problems in evolutionary immune network clustering, such as the lack of guidance in the clustering process, the sensitivity of the fuzzy boundary and the difficulty in determining parameters. To solve these problems, an artificial immune network clustering algorithm based on a cultural algorithm is proposed. Three kinds of knowledge are constructed: normative knowledge is used to standardize the spatial scope of population initialization, avoiding blindness; state knowledge is used to distinguish antigens and take immune defense measures to prevent noise and unclear network structure caused by boundary; topology knowledge is used to guide the optimal antibody search. The clone mutation operation of the traditional method is improved, and a compression threshold adaptive determination method is proposed based on the shadow sets theory. The experimental results show that the proposed method can effectively overcome the above problems, and the clustering performance on a synthetic dataset and an actual dataset is satisfactory.