Real estate practitioners have been struggling with accurate house price predictions for a long time. The traditional approaches lack when it comes to complex market patterns, which is why there has been a shift in recent years towards machine learning approaches. We evaluated various models such as XGBoost, LightGBM, Random Forest to name a few. Each tested model reflected various levels of strength, but also its weakness in many aspects. Some models were exceptionally accurate but not easy to interpret, while others are easy to understand but not as precise. The researchers decided to try something different: combining XGBoost and LightGBM with an ensemble hybrid model. The underlying idea was straightforward to bring the potential of both together. With real estate data of an Indian City, we processed our dataset, then trained several models to compare and benchmark, and tested everything with usual metrics such MAE, RMSE, and R2 scores. The results sparked curiosity. We observed that our ensemble approach consistently surpassed various individual models, showing better accuracy and sensitivity in different market conditions. It effectively addressed various problems like overfitting and high computational costs. In researching the previous research work, we noticed various big miss - most especially the challenge of balancing forecasting ability with interpretability. Our research and experiment suggest that ensemble learning offers a effective solution to real estate price prediction.