Automatic breast cancer classification technology in histopathology images plays a crucial role in accurate diagnosis and treatment planning. Recently, histopathology image classification approaches based on ResNet architecture have been widely proposed as they vastly improve the classification accuracy using skip connection to prevent vanishing gradient issues by combining low-level and high-level feature information. However, traditional ResNet architecture suffers from challenges such as data imbalance and interpretability, which still require cross-domain knowledge and collaboration among medical experts. This study proposes a new breast cancer classification method using a Dual-Activated Lightweight Attention ResNet50 (DALAResNet50) approach to address data imbalance and interpretability challenges effectively. The proposed approach combines a pre-trained deep ResNet50 method and a lightweight attention mechanism to accomplish the classification by embedding an attention module in layer 4 of ResNet50 and adding two fully connected layers based on LeakyReLU and ReLU activation functions to enhance the ability of feature learning. The proposed DALAResNet50 approach has been evaluated on breast cancer histopathology images derived from BreakHis Datase at magnification factors of 40X, 100X, 200X, and 400X, achieving accuracies of 98.5%, 98.7%, 97.9%, and 94.3%, respectively. The DALAResNet50 approach was comprehensively assessed with popular deep learning approaches such as SEResNet50, DensNet121, VGG16, VGG16Inception, ViT, Swin-Transformer, Dinov2_Vitb14, and ResNet50 models. The reported results of DALAResNet50 outperform the compared approaches regarding precision, accuracy, recall, F1 score, and GMean, demonstrating significant robustness and broad applicability when dealing with different magnifications and unbalanced breast cancer datasets