The Effect of Meta-heuristic Methods on the Performance of ImageClassification

Abstract

Classification accuracy is strongly affected by the quality of the input features. In recent years, datasets have increased in size and number of features. Analysis of huge datasets can be challenging due to redundant, noisy, and irrelevant features that may
reduce the classifier's performance. Feature selection is a vital process in which the best subset of features from the original dataset is chosen. The feature selection strategy is critical for increasing classification accuracy while decreasing computational costs. This research proposed a method for classifying lip print images by exploiting meta-heuristic methods and optimization-based feature selection methods. It involves four main phases: pre-processing, feature extraction, feature selection, and classification. After pre-processing, the features are extracted from the enhanced image. Meta-heuristic methods such as Genetic Algorithm (GA), Particle
Swarm Optimization (PSO), and Water Cycle Algorithm (WCA) are studied for feature selection using the mean function as the objective function. Finally, the lip print images are classified using a support vector machine (SVM). In this research, the experimental results are compared in terms of accuracy, error, sensitivity, and precision rate between three meta-heuristic methods and the accuracy rate of the proposed method with other algorithms that do not use meta-heuristic methods. The accuracy reached 97.9%, 96.8%, and 95% using WCA, PSO, and GA, respectively.