Cervical cancer remains one of the leading causes of mortality among women, which requires early detection and treatment to mitigate its impact. Recent advancements in medical image classification have demonstrated significant efficacy, with ensemble learning strategies playing a crucial role. Ensemble learning takes advantage of the combined strengths of multiple models to improve classification accuracy by creating a stronger and more accurate predictive model. This study presents an ensemble learning approach incorporating preprocessing techniques, image enhancement methods, and six diverse convolutional neural network (CNN) architectures for the classification of cervical cytology images from a Vietnamese dataset. Our ensemble models demonstrated superior classification performances across various metrics. Moreover, we observed a significant influence of image size variations on model efficacy, highlighting the importance of standardized image preprocessing.