Anemia, characterized by a deficiency of red blood cells or hemoglobin, is a prevalent condition with significant public health implications. This study aims to develop and compare the performance and interpretability of two machine learning models—Random Forest and Neural Networks—for the classification of anemia, employing LIME (Local Interpretable Model-agnostic Explanations) to enhance model transparency. The analysis was conducted under different conditions: handling class imbalance with and without SMOTE (Synthetic Minority Over-sampling Technique) and using two different feature selection methods, correlation and RFECV (Recursive Feature Elimination cross-validation). SMOTE was applied to address the class imbalance by creating synthetic samples from the minority class, while correlation and RFECV were used to select the most relevant features based on their correlation with the target variable and recursive elimination, respectively. Results: LIME-generated explanations provided valuable insights into the decision-making processes of both models, enhancing transparency by elucidating which features were most influential in predicting anemia. Comparing the models under different conditions highlighted their strengths and weaknesses in terms of performance and interpretability. The study underscores the significance of model explainability in healthcare applications. The LIME-based approach significantly enhances the interpretability of machine learning models, making them more transparent and trustworthy for anemia classification. These findings have broader implications for the development and deployment of machine-learning models in critical medical domains, where understanding the reasoning behind a model’s prediction is essential for clinical acceptance and utility. Integrating advanced machine learning techniques with interpretability tools like LIME can improve both performance and trustworthiness, leading to better patient outcomes and more informed clinical decisions.