Abstract
In this paper, particle swarm optimization is incorporated into an improved bacterial foraging optimization algorithm, which is applied to classifying imbalanced data to solve the problem of how original bacterial foraging optimization easily falls into local optimization. In this study, the borderline synthetic minority oversampling technique (Borderline-SMOTE) and Tomek link are used to pre-process imbalanced data. Then, the proposed algorithm is used to classify the imbalanced data. In the proposed algorithm, firstly, the chemotaxis process is improved. The particle swarm optimization (PSO) algorithm is used to search first and then treat the result as bacteria, improving the global searching ability of bacterial foraging optimization (BFO). Secondly, the reproduction operation is improved and the selection standard of survival of the cost is improved. Finally, we improve elimination and dispersal operation, and the population evolution factor is introduced to prevent the population from stagnating and falling into a local optimum. In this paper, three data sets are used to test the performance of the proposed algorithm. The simulation results show that the classification accuracy of the proposed algorithm is better than the existing approaches.
Highlights
In machine learning the imbalanced distribution of categories is called an imbalanced problem.When conventional algorithms are directly applied to this problem, the classification results tend to be biased towards most classes, resulting in a few classes not being correctly identified
Aiming at the BFO algorithm shortcoming of falling into a local optimum, we propose the incorporation of particle swarm optimization into an improved bacterial foraging optimization to solve these problems
In order to verify the performance of the proposed algorithm, ovarian cancer microarray data, a spam email dataset and a zoo dataset are used for simulation experiments
Summary
In machine learning the imbalanced distribution of categories is called an imbalanced problem.When conventional algorithms are directly applied to this problem, the classification results tend to be biased towards most classes, resulting in a few classes not being correctly identified. Various algorithms, including k nearest neighbor (KNN), decision tree (DT), artificial neural network (ANN), and the genetic algorithm (GA), have been recommended for data mining [10,11,12,13,14,15,16,17]. These algorithms usually assume that datasets are distributed evenly among different classes and that some classes may be ignored
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.