Chest X ray imaging is one of the most essential diagnostic tools present in modern health care. It does not require an invasive nature, and it is pretty cost-effective. It forms a readily available diagnostic procedure for most pulmonary and thoracic diseases. These could be useful in detecting pleural effusion, pneumonia, and cardiomegaly earlier and more accurately hence improving the outcome of treatment. The data captured in medical imaging is very complex, and the need for subtle patterns to be detected requires deep learning for extracting high-level features that allow for automatic and precise diagnosis. This paper is not to present a new model but optimize the performance of Convolutional Neural Networks (CNNs) through hyper parameter tuning. With a total of 5,606 chest X-ray images over 15 classes of diseases, the paper fine-tunes key hyper parameters: learning rate, batch size, and dropout rates using Grid Search and Auction Based Optimization Approaches (ABOA). Grid Search exhaustively explores the predefined parameter combinations, thereby ensuring optimal results, whereas ABOA applies a flexible, action-driven strategy to efficiently navigate the hyper parameter space, leading to better configurations. ABOA is not a fixed algorithm but a dynamic strategy that adapts to different optimization tasks, offering superior results in medical image classification. It performs better than traditional methods because it is more flexible and provides improved model accuracy and efficiency.