An Adaptive Support Vector Machine Optimized by an Improved Starfish Optimization Algorithm for Hyperspectral Image Classification

This study proposes an adaptive Support Vector Machine (SVM) classification method based on an enhanced Starfish Optimization Algorithm (SFOAE-SVM) for hyperspectral image (HSI) classification. HSI classification remains a critical challenge in remote sensing due to the high dimensionality of spectral features, spectral mixing, scarcity of labeled samples, and complex land-cover distributions. The SFOAE algorithm is used for global hyperparameter optimization of SVMs, accounting for the distributional characteristics of the target HSI data. The approach aims to improve search capability and reduce the likelihood of convergence to local optima by combining multi-dimensional topology-oriented expansion with global exploration. Experimental results demonstrate that SFOAE-SVM achieves competitive classification accuracy and stable performance compared with conventional SVM parameter selection strategies and other optimization-based methods across three benchmark hyperspectral remote-sensing datasets. These results indicate that the proposed method offers a promising optimization-assisted SVM framework for hyperspectral remote-sensing image classification.