Abstract
AbstractIn recent years, Artificial Immune Recognition System (AIRS) has attracted the attention of many artificial intelligence (AI) researchers and has been applied in different domains and applications. AIRS is an immune-inspired supervised learning algorithm known for its strong capacity in making decisions and solving real-world problems. Various AIRS versions have been proposed and have been proven to be effective in resolving difficult classification problems. Nevertheless, many issues are presented in these versions. For instance, existing AIRS methods require a huge number of default parameters which could have a negative impact on the decision making process. In addition, the majority of AIRS approaches use all the features during the learning and classification phase. However, many of these features could be irrelevant and redundant and could degrade the predictive model performance. Moreover, AIRS algorithms are unable to handle the uncertainty that could spread into the different stages of the classification process. To overcome all these issues and improve the classification efficiency, we propose in this paper a new optimized AIRS approach under an uncertain environment. In this approach, we opt for a feature selection method in order to reduce the number of features and select only the relevant ones. The main objective of our paper is to enhance the classification accuracy while handling uncertainty under the belief function theory. To do so, we employ two different optimization techniques, which are Genetic Algorithm (GA) and Gradient Descent method under an uncertain environment. Experiments executed on diverse real world data sets proved that our approach beats traditional AIRS versions in terms of classification accuracy.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have