Ensemble Learning for Breast Cancer Image Classification Using Pre-trained CNNs and Random Oversampling Method

The classification of breast cancer images presents significant challenges, especially when dealing with imbalanced datasets that underrepresent minority classes. This study tackles the issue by implementing ensemble learning techniques with pre-trained convolutional neural networks (CNNs) such as ResNet50, GoogleNet, and EfficientNet. The random oversampling method was employed to address the imbalance, which effectively increases the representation of minority classes within the dataset. This study performs detailed data analysis after this adjustment to improve the model’s ability to identify these minority classes. By combining the predictions of the three models into an ensemble, the approach aims to improve both the accuracy and robustness of the classification process. The ensemble method works much better at finding minority classes, as shown by the results of experiments. This proves that it can handle uneven datasets when classifying breast cancer images.