Chaos follow the leader algorithm: Application to data classification

Abstract

Classification is the process of systematically arranging the data into multiple groups, which gives a better understanding of large datasets. Considerable research has been carried out on data classification to develop a predictive model with high accuracy. This paper introduces chaos in the Follow the leader (FTL) optimization algorithm to achieve global optima for imbalanced data classification problems. In the FTL algorithm, the random parameter ω may not give enough exploration to the particles, which may lead to local stagnation problems. One of the solutions to this problem can be the replacement of the random parameter ω by a chaotic variable. For a detailed study on the chaotic behavior of ω in FTL, the algorithm is implemented with ten different chaotic maps and tested on twenty-four standard benchmark functions. Thereafter, the best chaos FTL (cFTL) algorithm is tested on six different complex real engineering problems and compared with ten meta-heuristic algorithms. Moreover, to further validate the performance of cFT, it has been applied to ten standard classification datasets collected from UCI and Kaggle data repositories. The obtained results show that cFTL outperforms FTL and other meta-heuristic algorithms.