Search

In this study, we proposed a new metaheuristic algorithm called Electrical Search Algorithm (ESA). The proposed algorithm is based on the movement of electricity in high-resistive areas such as wood, glass, and gases. ESA has a unique initialization scheme that only one agent initializes at the lower and upper bounds of the search space, which creates structures called poles. After that, ESA uses unique exploration and exploitation strategies to search. The search mechanism is based on electrons moving to opposite poles. ESA differs from other metaheuristics compared to its initialization scheme, pole search mechanism, and update strategy of the best solutions. ESA was tested with the "100-Digit Challenge" benchmark functions in the IEEE-CEC-2019, four well-known benchmark functions, and an np-hard clustering problem. For the clustering problem, we used four well-known datasets: Iris, Wine, Seeds, and Hepatitis C Virus. ESA was compared with seven different metaheuristic algorithms on these well-known benchmark functions, and the results of the clustering problem were compared with the K-Means algorithm. Additionally, Friedman Signed Rank and post hoc Wilcoxon Test were run to show the significance of the results. In all of the well-known benchmark functions, ESA either offered the best results or similar results to other compared algorithms. The score of the ESA on the IEEE-CEC-2019 benchmark functions shows us that even with the minor evaluation numbers, ESA can achieve similar results to the competing algorithms. Results show that ESA has a robust mechanism for not trapping in local points and moves slow but persistent rate. [ABSTRACT FROM AUTHOR]

Feature transformation methods can be used effectively to improve the performance of the machine learning techniques, such as classification and clustering, via enhancing the discrimination among the samples. Additionally, binary representation of features may be necessary in several circumstances. For example, distributed systems may have limited bandwidth, storage, and energy that necessitate rough quantization of the measurements. This paper proposes a random projection-based feature transformation method that maps the data points from the original space into a new binary space. In the proposed transformation method, the random projection process is formulated as an optimization problem. A modified chaotic version of the sine–cosine algorithm is used to find the optimal random projector. The performance of the proposed transformation method is validated using 15 standard UCI benchmark datasets against the linear discernment analysis. Moreover, it is compared to other metaheuristic-based feature reduction methods including the gray wolf optimizer, the genetic algorithm, and the antlion optimizer. The obtained results assure the superiority of the proposed transformation method in terms of classification accuracy and storage requirements. [ABSTRACT FROM AUTHOR]