Negative Selection Based Data Classification with Flexible Boundaries

One of the most important artificial immune algorithms is negative selection algorithm, which is an anomaly detection and pattern recognition technique; however, recent research has shown the successful application of this algorithm in data classification. Most of the negative selection methods consider deterministic boundaries to distinguish between self and non-self-spaces. In this paper, two negative selection based algorithms are proposed for two-class and multi-class classification problems; using a Gaussian mixture model which is fitted on normal space to create a flexible boundary between self and non-self-spaces, by determining the dynamic subsets of effective detectors to solve the problem of data classification. Initialization of effective parameters such as the detection threshold, the maximum number of detectors etc. for each dataset, is one of the challenges in negative selection based classification algorithms, which affects the precision and accuracy of the classification; therefore, an adaptive and optimal calculation of these parameters is necessary. To overcome this problem, the particle swarm optimization algorithm has been used to properly set the parameters of the proposed methods. The experimental results showed that using a Gaussian mixture model and dynamic adjustment of parameters such as optimum number of Gaussian components according to the shape of the boundaries, creation of appropriate number of detectors, and also automatic adjustment of the training and testing thresholds, using particle swarm optimization algorithm as well as utilization of a combinatorial objective function has led to a better classification with fewer detectors. The proposed algorithms showed competitive performance compared with some of the existing classification algorithms, including several immune-inspired models, especially negative selection ones, and other traditional classification methods.