Enhancing Multi-Class Lung Disease Classification with Bagging-Based Deep Learning Ensembles: A Comparative Study of Five Architectures

Accurate classification of multi-class lung diseases is very important and crucial for proper diagnosis and treatment in medical imaging. This paper presents evaluation of the ensemble learning method using bagging to improve performance of deep learning models on that important task. Five powerful CNN architectures, ResNet50, VGG16, VGG19, MobileNetV2 and DenseNet201, have been analyzed using public sources of lung disease images. Classification accuracy of each model has also been individually trained and tested. To enhance performance, a bagging ensemble approach was used, where the predictions of the models were combined to exploit their complementary strengths. The overall accuracy was used as the evaluation metric to assess the performance of both individual models and the ensemble strategy. The results show that the bagging ensemble greatly enhances classification accuracy, reaching an overall accuracy of 96.63%. Individually, the CNN models achieved accuracies ranging from 89.19% to 94.48%. This analysis highlights the strength and efficiency of ensemble learning in multi-class lung disease classification tasks.