The recurrence of thyroid cancer has been elevating considerably, with women having a higher prevalence than men. Despite breakthroughs in diagnostic tools, the likelihood of recurrence is a considerable issue. Predicting early thyroid cancer recurrence is crucial for patient outcomes, where Machine Learning can significantly enhance accuracy. We utilize a recent dataset comprising 13 clinicopathological features of 383 patient records. We train and test several mainstream Machine Learning algorithms such as Logistic Regression, XGBoost, Decision Tree, Support Vector Machine, and Random Forest. In addition, we introduce a novel way of selecting features by employing multiple feature selection techniques, which contribute to improving the model’s overall effectiveness by reducing the number of features in the training dataset. Thus, leveraging our custom training pipeline, we trained our models and found them to outperform the same models’ performance in the recent existing studies. Among all models, the Random Forest demonstrated a superior accuracy of 98.70% for predicting thyroid cancer recurrence. Thus, our study demonstrates the utility of our custom training pipeline of models incorporating advanced Machine Learning techniques in predicting thyroid cancer recurrence accurately, which will undoubtedly contribute to developing personalized treatment strategies and improving patient outcomes.