The selection of an optimal Controlled Ovarian Stimulation (COS) protocol is critical in in-vitro fertilization (IVF) to minimize total fertilization failure and enhance treatment success. This study introduces a data-driven framework that leverages prescriptive analytics and machine learning to personalize COS recommendations, thus improving fertilization results. Using electronic medical records (EMRs) from 15,193 IVF cycles at Assuta Ramat Hahayal Medical Center, we applied K-means clustering to group patients based on pre-treatment patient characteristics and categorized COS protocols according to treatment attributes. Predictive models were then developed for each patient-treatment cluster combination to estimate the likelihood of fertilization success under different COS regimens. By integrating causal inference techniques, the algorithm identifies the COS protocol with the highest predicted success rate for each patient, enabling data-informed decision-making. A retrospective evaluation on 11,947 IVF cycles showed that among the 3,326 cycles (27.8%) that followed the model’s recommendations, the total fertilization failure (TFF) rate was 19.3%, compared to 36.8% across all cycles. This corresponds to a 17% absolute reduction in TFF relative to standard clinical practice. Given that total fertilization failure prematurely terminates the IVF cycle, necessitating additional costly and emotionally taxing treatment attempts, this framework provides a crucial decision-support tool to improve patient outcomes. By systematically optimizing COS selection, this prescriptive analytics approach has the potential to significantly enhance IVF success rates. Furthermore, the methodology can be extended to other key decision points in the IVF process, contributing to more effective and personalized fertility treatments.