<![CDATA[In the development of artificial intelligence technology, particularly deep learning, convolutional neural networks (CNNs) have become one of the most popular architectures for image classification tasks. The ability of CNN to extract features from image data without manual processing makes it superior, especially in handling complex image data. One of the important components in CNN is pooling. Which serves to reduce the dimensions of the data while preserving important information. This study analyzes the impact of the combination of local and global pooling on the performance of CNN in classifying the CIFAR-10. This approach was carried out by training two CNN models, namely a model with a combination of pooling and a model with local pooling only. The training process uses the k-fold cross validation method. Performance evaluation was conducted using accuracy, precision, recall, and F1-score metrics. The research results show that the model with the pooling combination achieved an average accuracy of 83.75%, slightly higher than the local pooling, which resulted in an accuracy of 83.61%. Additionally, the model with the pooling combination demonstrated stability during training and good generalization capability. This research contributes to the optimization of CNN architecture by demonstrating that the combination of local and global pooling has the potential to improve model performance, especially on datasets with high feature diversity.]]>
