Elitist Binary Wolf Search Algorithm for Heuristic Feature Selection in High-Dimensional Bioinformatics Datasets

Jinyan Li,Simon Fong,Richard Millham,Kelvin K. L. Wong,Raymond K. Wong

doi:10.1038/s41598-017-04037-5

Abstract

Due to the high-dimensional characteristics of dataset, we propose a new method based on the Wolf Search Algorithm (WSA) for optimising the feature selection problem. The proposed approach uses the natural strategy established by Charles Darwin; that is, ‘It is not the strongest of the species that survives, but the most adaptable’. This means that in the evolution of a swarm, the elitists are motivated to quickly obtain more and better resources. The memory function helps the proposed method to avoid repeat searches for the worst position in order to enhance the effectiveness of the search, while the binary strategy simplifies the feature selection problem into a similar problem of function optimisation. Furthermore, the wrapper strategy gathers these strengthened wolves with the classifier of extreme learning machine to find a sub-dataset with a reasonable number of features that offers the maximum correctness of global classification models. The experimental results from the six public high-dimensional bioinformatics datasets tested demonstrate that the proposed method can best some of the conventional feature selection methods up to 29% in classification accuracy, and outperform previous WSAs by up to 99.81% in computational time.

Highlights

Feature selection is a commonly and effectively used method of feature dimension reduction in selecting a suitable low-dimensional sub-dataset from an initial high-dimensional dataset[4, 5]
Compared with other non-iterative traditional methods, such as Binary Particle Swarm Optimisation (BPSO) and Wolf Search Algorithm (WSA), the proposed method could improve the accuracy of classification models with high computational and convergence speeds
To maintain the fairness of the experiment, because our proposed method, particle swarm optimisation (PSO), BPSO and WSA are random searching strategy algorithms, their experiments are repeated ten times, and the final results are used as the mean value

Summary

Introduction

Feature selection is a commonly and effectively used method of feature dimension reduction in selecting a suitable low-dimensional sub-dataset from an initial high-dimensional dataset[4, 5]. × n!] candidate solutions to select n sub-features from N features in the search space. WSAs have adopted wrapper strategies with different traditional classifiers, significantly exceeding some well-known swarm-based feature selection methods in classification performances[11, 12]. The new binary version of the WSA for feature selection proposed in this paper is called the Elitist Binary Wolf Search Algorithm (EBWSA). The size of the search space is 2 × N, where N denotes the number of features In such cases, the program avoids simultaneously considering the best length and combination of sub-features from the 2N − 1 possible candidate solutions. Subset generation is a search process that uses corresponding strategies to select preselected sub-datasets. Feature selection approaches can be divided into two types: those that function according to the strategy for searching subsets, and those that do so according to the evaluation standard for features

Methods

Results

Conclusion