Recognizing emotional states in domestic cats presents significant challenges due to the subtle nature of feline facial expressions and behavioral cues. This research introduces a robust deep learning framework for automated feline emotion recognition across seven distinct emotional categories: angry, disgusted, happy, normal, sad, scared, and surprised. Leveraging the EfficientNet-B0 architecture with transfer learning, our model achieves exceptional classification performance with 99.20% training accuracy and 99.67% validation accuracy on a comprehensive dataset of feline facial expressions. The methodology incorporates strategic data partitioning, advanced augmentation techniques, and weighted random sampling to address class imbalance. Additionally, comprehensive statistical validation through parametric and non-parametric tests confirms the robustness and generalizability of our approach. Experimental results demonstrate near-perfect classification metrics across all emotional categories, with precision, recall, and F1-scores of 0.99 or higher. This research advances the field of animal-computer interaction by providing veterinarians, animal behaviorists, and pet owners with an accurate, automated tool for understanding feline emotional states, thereby contributing to improved animal welfare and enhanced human-feline relationships.