Flight Height Whale Optimization Algorithm with Convolution Neural Network for Imbalanced Data Classification

Data sampling is a statistical analysis method used to select, manipulate and analyze a representative subset of data points to identify the patterns and trends in huge datasets.This approach addresses the problem of imbalanced classification, particularly with severely imbalanced class distributions through data preparation methods.Various Flight Height based Whale Optimization Algorithms (FHWOA) combined with Convolution Neural Networks (CNN) are proposed for imbalanced data classification.By simulating various flight heights, the FHWOA enhances the exploration and exploitation abilities, thereby optimally handling imbalanced datasets.In FHWOA, the varying exploration altitudes function as flight heights, which facilitate better search space navigation for selecting optimal features to improve imbalanced dataset classification.Additionally, FHWOA dynamically adapts to diverse search spaces, ensuring optimal feature selection.The CNN automatically learns hierarchical spatial features, enabling efficient feature extraction and enhancing model performance in imbalanced datasets.The performance of the FHWOA-CNN is measured in terms of accuracy, f1-score, recall, precision and Area Under the Curve (AUC).The FHWOA-CNN achieves accuracies of 99.85%, 97.50%, 99.94% and 98.98% on the Breast, Central Nervous System (CNS), Leukemia and Lung datasets, respectively, thereby outperforming Synthetic Minority Oversampling Technique-Normalization-based CNN (SMOTE-Norm-CNN).