This paper presents a novel approach to Electronic Health Record (EHR) analysis using machine learning (ML) models for phenotyping, aimed at enhancing the recognition and treatment of Anemia. Utilizing the MIMIC-III dataset, the study involves data pre-processing and subject data analysis to uncover key insights into the prevalence of anemia, gender distribution, comorbidities, and Intensive Care Unit (ICU) stays. Clustering algorithms such as K-Means and HDBSCAN, leveraging Levenshtein Distance and TF-IDF Vectorizer metrics, produce distinct patient clusters characterized by unique features and comorbidities. Evaluation through Length of Stay (LoS) estimation demonstrates the effectiveness of the proposed algorithms in improving predictive accuracy, with HDBSCAN on the Vectorizer object achieving a 50.82% reduction in Root Mean Square Error (RMSE) compared to traditional methods. This data-driven approach shows significant potential for personalized treatment strategies and enhanced healthcare outcomes, particularly across diverse healthcare settings.