Hummingbird Optimization with Deep Learning Enabled Feature Reduction and Classification Approach for High Dimensional Data

This paper proposes a novel technique for image classification tasks that handles high-dimensional data: Hummingbird Optimization Deep Learning-based Feature Reduction and Classification (HBODL-FRC). The proposed method incorporates several cutting-edge methodologies to improve the precision and effectiveness of the categorization process. To ensure the quality of the input data, bilateral filtering is first used to successfully remove noise. After that, key features are extracted by combining texture features and histograms, which form the basis of the feature reduction procedure that follows. One of the main innovations in this work is the deployment of the Hummingbird Optimization (HBO) algorithm, which is used to intentionally minimize the feature sets’ complexity. The final classification of the images is accomplished by feeding the optimized feature set into a “multi-head attention-based bidirectional long short-term memory” (MABi-LSTM) model. To further improve classification accuracy, the MABi-LSTM model makes use of the attention mechanism to concentrate on the most pertinent portions of the data. Using benchmark datasets, the HBODL-FRC model’s performance was thoroughly assessed and compared to the existing techniques. The HBODL-FRC model performs better than previous methods in terms of robustness and classification accuracy, according to extensive experimental results.