This study evaluates the efficacy of various machine learning optimizers for predicting keratoconus progression using the Harvard Keratoconus Dataset, which includes comprehensive clinical data from diagnosed patients. The research utilized a regression model with three dense layers, applying rigorous preprocessing techniques like data normalization and cleaning to ensure consistency and quality. Our comparative analysis revealed that the Adam optimizer outperformed others by achieving faster convergence and maintaining lower validation errors. This superiority is attributed to Adam’s adaptive learning rate mechanism, which effectively manages the high variability typical of clinical data in keratoconus. The findings demonstrate the importance of optimizer selection in enhancing the predictive accuracy of machine learning models in medical applications. Adam’s robust performance underscores its potential for integration into real-time clinical systems for monitoring keratoconus progression, thereby improving patient management and treatment outcomes. This study contributes valuable insights into the application of machine learning in ophthalmology and sets a precedent for future research in predictive medical modeling.