<![CDATA[Advances in artificial intelligence are driving the optimization of deep learning models for image analysis. Air pollution is a significant environmental problem that impacts human health and the balance of ecosystems. Increased emissions from industrial activities, transportation, and the burning of fossil fuels are major factors contributing to deteriorating air quality in various regions. This study aims to analyze the influence of data preprocessing and training strategies on model performance, including the removal of duplicate image data, data splitting, data augmentation, the use of class weights on training data, and learning rate tuning using the Inception V3 transfer learning model. The results show that the combination of these strategies achieved an accuracy of 89.28% with an error rate of only 0.338, demonstrating stability and good model generalization capabilities. The application of appropriate and effective preprocessing and training strategies enhances model performance. Consequently, a more accurate and reliable model can support data-driven decision-making more efficiently and contribute to various sectors such as health and the environment in supporting sustainable development.]]>
Copyrights © 2026
