A Review of House Price Regression Prediction Methods

Abstract

Real estate is an important sector of the national economy. Fluctuations in housing prices affect residents' lives, financial security and the overall economy. Accurate prediction of housing prices can provide support for market entities' decision-making and government housing regulation. With the development of econometrics and machine learning, housing price regression prediction has gradually evolved from traditional characteristic price models to ensemble learning. This paper classifies existing prediction methods into traditional econometric methods, classical machine learning, hybrid and deep learning. It analyzes the principles, applicable scenarios and advantages and disadvantages of each method, and summarizes empirical conclusions from feature engineering - multi-source data fusion and the model applicability in different market environments. The results show that tree ensemble models such as random forest and XGBoost generally have better prediction effects than traditional linear regression. Hybrid models and multi-source data fusion further enhance the ability to predict housing prices. Currently, there are still problems in this field such as limited data quality, weak model interpretability, and difficulty in modeling temporal and spatial dynamic relationships. Future research can focus on exploring housing price prediction under causal inference, the combination of spatial econometrics and machine learning, and non-structured data mining, etc.