The field of hard milling has recently witnessed growing interest in environmentally sustainable machining practices. Among these, Minimum Quantity Lubrication (MQL) has emerged as an effective strategy, offering not only reduced environmental impact but also economic benefits and enhanced cooling performance compared to conventional flood cooling methods. In hard milling operations, cutting temperature is a critical factor that significantly influences the quality of the finished component. Proper control of this parameter is essential for producing high-precision workpieces, yet measuring cutting temperature is often complex, time-consuming, and costly. These challenges can be effectively addressed by predicting cutting temperature using advanced Machine Learning (ML) models, which offer a faster and more efficient alternative to direct measurement. In this context, the present study investigates and compares the performance of Conventional Minimum Quantity Lubrication (CMQL) and Graphene-Enhanced MQL (GEMQL), with sesame oil serving as the base fluid, in terms of their effect on cutting temperature. The experiments are structured using a Taguchi L36 orthogonal array, with key variables including cutting speed, feed rate, MQL jet pressure, and the type of cooling applied. Additionally, the study explores the predictive capabilities of various advanced ML models, including Decision Tree, XGBoost Regressor, K-Nearest Neighbor, Random Forest Regressor, and CatBoost Regressor, along with a Hybrid Stacking Machine Learning Model (HSMLM) for estimating cutting temperature. The results demonstrate that the GEMQL setup reduced cutting temperature by 36.8% compared to the CMQL environment. Among all the ML models tested, HSMLM exhibited superior predictive performance, achieving the best evaluation metrics with a mean absolute error of 3.15, root mean squared error (RMSE) of 5.3, mean absolute percentage error of 3.9, coefficient of determination (R2) of 0.91, and an overall accuracy of 96%.