A modified whale optimization algorithm for enhancing the features selection process in machine learning

Syed, Ezaz Uddin and Masood, Mohsin and Fouad, Mohamed Mostafa and Glesk, Ivan (2021) A modified whale optimization algorithm for enhancing the features selection process in machine learning. In: 29th Telecommunications Forum TELFOR 2021, 2021-11-23 - 2021-11-24, School of Electrical Engineering, University of Belgrade.

[thumbnail of MWOA-ML_submitted] Text. Filename: MWOA_ML_submitted.pdf
Restricted to Registered users only

Download (1MB) | Request a copy
[thumbnail of Syed-etal-TELFOR-2021-A-modified-whale-optimization-algorithm-for-enhancing-the-features-selection]
Text. Filename: Syed_etal_TELFOR_2021_A_modified_whale_optimization_algorithm_for_enhancing_the_features_selection.pdf
Accepted Author Manuscript

Download (272kB)| Preview


In recent years, when there is an abundance of large datasets in various fields, the importance of feature selection problem has become critical for researchers. The real world applications rely on large datasets, which implies that datasets have hundreds of instances and attributes. Finding a better way of optimum feature selection could significantly improve the machine learning predictions. Recently, metaheuristics have gained momentous popularity for solving feature selection problem. Whale Optimization Algorithm has gained significant attention by the researcher community searching to solve the feature selection problem. However, the exploration problem in whale optimization algorithm still exists and remains to be researched as various parameters within the whale algorithm have been ignored and not introduced into machine learning models. This paper proposes a new and improved version of the whale algorithm entitled Modified Whale Optimization Algorithm (MWOA) that hybrid with the machine learning models such as logistic regression, decision tree, random forest, K-nearest neighbour, support vector machine, naïve Bayes model. To test this new approach and the performance, the breast cancer datasets were used for MWOA evaluation. The test results revealed the superiority of this model when compared to the results obtained by machine learning models.


Syed, Ezaz Uddin, Masood, Mohsin ORCID logoORCID: https://orcid.org/0000-0003-4388-569X, Fouad, Mohamed Mostafa and Glesk, Ivan ORCID logoORCID: https://orcid.org/0000-0002-3176-8069;