Classroom behavior detection is a significant research direction in developing intelligent education. Due to challenges such as small objects, occlusions, and cluttered scenes in classroom environments, conventional object detection algorithms often suffer from low accuracy and poor real-time performance. To address these issues, this paper proposes an improved YOLO11n-based algorithm, YOLO-CBD. The YOLOv11 network structure is optimized by incorporating attention mechanisms and multi-scale feature fusion strategies. The Iterative Attentional Feature Fusion (iAFF) module is introduced to replace the C3k2 module, enhancing the ability of the model to detect small objects and handle occlusions. Additionally, a Hierarchical Feature Fusion (HFF) Block is integrated to capture multi-scale contextual information in classroom behavior detection tasks. The loss function is optimized better to adapt the model to classroom behavior detection scenarios. Experimental results demonstrate that, compared to the baseline model, the YOLO-CBD method achieves improved detection accuracy with only a slight increase in parameters while maintaining detection speed. Furthermore, YOLO-CBD outperforms other general object detection models in classroom behavior detection tasks.