A MACHINE LEARNING-BASED HYBRID FRAMEWORK FOR REAL ESTATE PRICE PREDICTION

Abstract

Accurate prediction of real estate prices plays a significant role in financial planning, investment decision-making, urban development, and smart city management. Traditional property valuation approaches mainly rely on statistical regression techniques and expert judgment, which often fail to model nonlinear market behavior and heterogeneous housing datasets. To overcome these limitations, this research proposes a novel Hybrid Adaptive Real Estate Prediction Algorithm (HAREPA) within a machine learning–based hybrid framework.

The proposed system integrates Linear Regression, K-Nearest Neighbor, and Support Vector Regression models through adaptive ensemble learning. Feature engineering techniques are employed to extract meaningful attributes and improve learning efficiency. Adaptive weights are automatically computed using inverse error analysis, enabling dynamic contribution of each model during prediction.

Experimental evaluation is conducted using a real-world real estate dataset obtained from Kaggle. Performance comparison using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and prediction accuracy demonstrates that the proposed HAREPA model significantly outperforms individual machine learning algorithms. The results confirm improved prediction accuracy, reduced error variance, and enhanced generalization capability.