International Journal of Electrical and Computer Engineering (IJECE) Vol. 12, No. 4, August 2022, pp. 4315~4326 ISSN: 2088-8708, DOI: 10.11591/ijece.v12i4.pp4315-4326  4315 Journal homepage: http://ijece.iaescore.com Cuckoo algorithm with great deluge local-search for feature selection problems Mutasem Khalil Alsmadi 1 , Malek Alzaqebah 2,3 , Sana Jawarneh 4 , Sami Brini 2,3 , Ibrahim Al-Marashdeh 1 , Khaoula Briki 2,3 , Nashat Alrefai 2,3 , Fahad Ali Alghamdi 1 , Maen T. Al-Rashdan 5 1 Department of Management Information Systems, College of Applied Studies and Community Service, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia 2 Department of Mathematics, College of Science, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia 3 Basic and Applied Scientific Research Center, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia 4 Computer Science Department, The Applied College, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia 5 Faculty of Ccience and Information Technology, Jadara University, Irbid, Jordan Article Info ABSTRACT Article history: Received Oct 16, 2020 Revised Mar 25, 2022 Accepted Apr 18, 2022 Feature selection problem is concerned with searching in a dataset for a set of features aiming to reduce the training time and enhance the accuracy of a classification method. Therefore, feature selection algorithms are proposed to choose important features from large and complex datasets. The cuckoo search (CS) algorithm is a type of natural-inspired optimization algorithms and is widely implemented to find the optimum solution for a specified problem. In this work, the cuckoo search algorithm is hybridized with a local search algorithm to find a satisfactory solution for the problem of feature selection. The great deluge (GD) algorithm is an iterative search procedure, that can accept some worse moves to find better solutions for the problem, also to increase the exploitation ability of CS. The comparison is also provided to examine the performance of the proposed method and the original CS algorithm. As result, using the UCI datasets the proposed algorithm outperforms the original algorithm and produces comparable results compared with some of the results from the literature. Keywords: Classification Cuckoo search Feature selection Great deluge Metaheuristic optimization This is an open access article under the CC BY-SA license. Corresponding Author: Mutasem Khalil Alsmadi Department of MIS, College of Applied Studies and Community Service, Imam Abdulrahman Bin Faisal University, P.O. Box 1982, 31441, City of Dammam, Saudi Arabia Email: mkalsmadi@iau.edu.sa 1. INTRODUCTION Classification is considered as one of the machine learning tasks, which have been widely used recently to categorize the data into classes [1]–[5]. Classification techniques predict the classes of the data instances based on a given set of data fields (features). Using the original number of features may be time- consuming and may mislead the classification process, so the methods for feature selection chose a minimum set of features that lead to better learning accuracy and less computational cost. The methods for feature selection are separated into three categories [6]: i) filter-based methods that use statistical approaches to assess the correlation between features and the class, ii) wrapper-based methods: assess the selected features’ subset using a machine learning algorithm, and iii) embedded-based methods: combines the advantage of wrapper methods and filter-based methods [6]–[8]. Feature selection is an nondeterministic polynomial (NP)-problem because of its high-dimensional space [9]–[12] where the exhaustive search is unfeasible. To perform the feature selection task, an efficient search algorithm is required. Swarm intelligence is a group of population-based algorithms that contain