Revised Artificial Immune Recognition System

Abstract

Artificial Immune Recognition System is a widely used bio-inspired algorithm that describes the recognition tasks of antigen by memory cells. Despite the success of the Artificial Immune Recognition System, the basic version has some drawbacks which have a direct impact on system efficiency in terms of the quality of the results, data explosion, and calculation cost. This paper investigates these disadvantages and proposes several modifications in the original version to overcome these problems. First, the concept of weight and lifetime counter was introduced for each memory cell to improve quality; second, a new mechanism was added to eliminate inactive memory cell models to reduce data explosion, and third, the structure of the memory cells set was replaced by a binary search tree to reduce processing time. Furthermore, this paper improves some algorithm functionalities especially in the mutation function and the memory cell introduction mechanism. The experimental results conducted on eleven public datasets show that the proposed method outperforms the original version, all the revised versions, and achieved a good rank compared to the other state-of-the-art methods with an average accuracy of 93.20 % on all tested datasets.

Keywords

• AIRS
• bio-inspired
• KD-tree
• kNN
• supervised learning