A cooperative swarm intelligence algorithm based on quantum-inspired and rough sets for feature selection. (January 2018)

• Record Type: Journal Article
• Authors: Zouache, Djaafar; Ben Abdelaziz, Fouad
• Is Part Of: Computers & industrial engineering. Volume 115(2018)
• Page Start: 26
• Page End: 36
• Publication Date: 2018-01
• Subjects: Feature selection -- Classification -- Rough sets -- Quantum computation -- Swarm intelligence; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2017.10.025
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: We propose a cooperative swarm intelligence algorithm for feature selection. It combines a firefly algorithm and a particle swarm optimization. It uses the quantum computing for the representation of the feature selection. It uses rough set theory to assess the relevance of feature subsets generated. It provides a better rate of feature reduction and a high accuracy classification. Abstract: Feature selection is an important preprocessing step for classification as it improves the accuracy and overcomes the complexity of the classification process. However, in order to find a potentially optimal feature subset for the feature selection problem, it is necessary to design an efficient exploration approach that can explore an enormous number of possible feature subsets. It is also necessary to use a powerful evaluation approach to assess the relevance of these feature subsets. This paper presents a new cooperative swarm intelligence algorithm for feature selection based on quantum computation and a combination of Firefly Algorithm (FA) and Particle Swarm Optimization (PSO). Quantum computation ensures a good trade-off between the exploration and the exploitation of the search space while the combination of the FA and PSO enables an effective exploration of all the possible feature subsets. We use rough set theory to assess the relevance of the potential generated feature subsets. We tested the proposed algorithm on eleven UCI datasets and compared with a deterministic rough set reduction algorithms and other swarm intelligence algorithms. The experiment results show clearly that our algorithm provides a better rate of feature reduction and a high accuracy classification.