A hybrid LSTM-GRU framework for lung cancer classification using GWO-WOA algorithm for hyperparameter tuning and BPSO...

Area of Science:

• Computational Biology
• Machine Learning in Healthcare
• Bioinformatics

Background:

• Early lung cancer detection is crucial for improving patient outcomes.
• Questionnaire data offers a low-cost, non-invasive method for initial screening.
• Traditional classifiers struggle with noisy, imbalanced, and redundant features in such datasets.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning framework for enhanced lung cancer classification.
• To optimize the model using metaheuristic algorithms for feature selection and hyperparameter tuning.
• To assess the model's performance on diverse public lung cancer datasets.

Main Methods:

• A hybrid Long Short-Term Memory (LSTM)-Gated Recurrent Unit (GRU) network was developed.
• Grey Wolf-Whale Optimization (GWO-WOA) algorithm optimized hyperparameters (learning rate, hidden units, layer depth).
• Binary Particle Swarm Optimization (BPSO) performed feature selection on preprocessed Kaggle lung cancer datasets.

Main Results:

• The GWO-WOA-LSTM-GRU model achieved 100.00% accuracy, precision, recall, and F1-score on a 309-sample dataset.
• On a 3000-sample dataset, the model reached 99.33% accuracy and F1-score.
• Outperformed single models (LSTM, GRU, CNN, SVM) with accuracies ranging from 77.42% to 98.33%.

Conclusions:

• The proposed hybrid deep learning model demonstrates superior performance and robustness in lung cancer classification.
• Metaheuristic optimization significantly enhances the accuracy and generalization of recurrent neural networks.
• This approach offers a reliable tool for early lung cancer detection and clinical risk stratification.