Data clustering using Clonal-Negative Selection Algorithm improved with chaotic function

Artificial immune systems (AIS) are a soft computing method that inspired from human immune system. Major usages of AIS are optimization and data mining applications as classification and clustering. In this paper we study a clustering problem and combine chaotic function with negative selection algorithm (NSA) to cover all of the problem space with the detector initial population and avoiding overlap between these detectors. Then in identification step we used clonal selection algorithm (CSA) for samples that NSA’s detectors can’t cluster, and with affinity measure the nearest cluster is calculated to determine each sample is a self (antibody) or non self (antigen) for a cluster. As a result we decreased number of detectors and increased accuracy in comparison with other methods that used NSA. Accuracy of clustering the Iris dataset with this method is 98.09.

کلیدواژه ها:

data mining ، artificial immune systems ، negative selection algorithm ، clonal selection algorithm ، chaos theory