Machine Learning Models for Predicting Residential Property Prices: An Intelligent and Robust Approach Based on XGBoost

Accurately forecasting residential property values is complicated in the property industry. Real estate markets are influenced by a wide range of elements that affect market behaviour, including economic, social, and geographical considerations. Due to the regional financing initiatives and the growing demand for living space, a remarkable increase in construction activities has increased in recent years. These changes affect the availability of living space and new factors that influence price trends. Therefore, a comprehensive understanding of environmental and market data is essential for developing a reliable and accurate model for real estate forecasting. In this work, we utilise XGBoost to predict price trends, a model renowned for accuracy, scalability, and interpretability. We optimise the model output through advanced feature engineering and hyperparameter tuning techniques. We then compare the predictions of several models, including Random Forest, Decision Tree, Support Vector Regression, Gradient Boosting Machines and Neural Networks. The results reveal that XGBoost outperforms these models and achieves superior accuracy in forecasting real estate prices. Key factors influencing housing costs, including Energy Performance Certificate (EPC) characteristics, location, and historical price trends, are discovered and studied. The dataset includes EPC features, historical transaction data and socio-economic indicators and offers a comprehensive basis for analysis. Our results underline the decisive role of EPC and location-based characteristics in determining property values and provide implementable knowledge for representatives of interest in the real estate sector. The model demonstrates excellent performance with low MAE and high R2 values, indicating its robustness and reliability. This study not only promotes the area of machine learning in real estate analysis but also offers a scalable model for collecting data and the prediction of real estate prices in other regions with comparable market dynamics. By overcoming challenges such as data shortages and data heterogeneity, our work has set a new benchmark for the accuracy and interpretability of real estate price forecasts.