Bat algorithm optimisation technique for feature selection on different dimension of datasets

Abstract

The advanced of Information Technology has resulting in the generation of numerous datasets with different dimensions. However, dealing with multi-dimensional datasets which typically contain large number of attributes, p has cause problems to classification process. Classifying different dimensional numerical data is a difficult problem as dealing with various feature spaces, could cause the performance of supervised learning method to suffer from the curse of dimensionality.  This condition eventually degrades both classification accuracy and efficiency. In a nutshell, not all attributes in the dataset can be used in the classification process since some features may lead to low performance of classifier. Feature selection (FS) is a good mechanism that minimises the dimensions of high-dimensional datasets and solve classification problems. This paper proposed Bat Algorithm (BA) for FS that were trained using a Support Vector Machine (SVM) classifier. The proposed algorithm was tested on six public datasets with different sizes and compared with other benchmark algorithms, such as Particle Swarm Optimisation (PSO) and Genetic Algorithm (GA). The experimental results indicated that the BA has outperformed the other two algorithms. In addition, the comparison details showed that binary BA is more competitive in terms of accuracy and the number of features when assessed on datasets with different sizes.