[Paper Review] Housing Market Prediction Problem using Different Machine Learning Algorithms: A Case Study

Shashi Bhushan Jha, Radu F. Babiceanu+2 more|arXiv (Cornell University)|Jun 17, 2020

Housing Market and Economics27 references19 citations

TL;DR

This study evaluates multiple machine learning algorithms—XGBoost, CatBoost, Random Forest, Lasso, and Voting Regressor—for predicting housing prices using a public dataset of 62,723 real estate records from Volusia County, Florida (2015–2019). XGBoost outperformed all other models in prediction accuracy, as measured by R², MSE, MAE, and computational efficiency, making it the optimal choice for housing market forecasting.

ABSTRACT

Developing an accurate prediction model for housing prices is always needed for socio-economic development and well-being of citizens. In this paper, a diverse set of machine learning algorithms such as XGBoost, CatBoost, Random Forest, Lasso, Voting Regressor, and others, are being employed to predict the housing prices using public available datasets. The housing datasets of 62,723 records from January 2015 to November 2019 are obtained from Florida Volusia County Property Appraiser website. The records are publicly available and include the real estate or economic database, maps, and other associated information. The database is usually updated weekly according to the State of Florida regulations. Then, the housing price prediction models using machine learning techniques are developed and their regression model performances are compared. Finally, an improved housing price prediction model for assisting the housing market is proposed. Particularly, a house seller or buyer, or a real estate broker can get insight in making better-informed decisions considering the housing price prediction. The empirical results illustrate that based on prediction model performance, Coefficient of Determination (R2), Mean Square Error (MSE), Mean Absolute Error (MAE), and computational time, the XGBoost algorithm performs superior to the other models to predict the housing price.

Motivation & Objective

• To develop and compare multiple machine learning models for accurate housing price prediction.
• To evaluate model performance using standard regression metrics such as R², MSE, and MAE.
• To identify the most effective algorithm for real-world housing market forecasting.
• To provide actionable insights for homebuyers, sellers, and real estate professionals.
• To contribute a robust, data-driven model for socio-economic planning and real estate decision-making.

Proposed method

• The study uses a public dataset of 62,723 real estate records from the Volusia County Property Appraiser website, covering January 2015 to November 2019.
• Features include property characteristics, location data, and economic indicators, with the target variable being the sale price.
• Six machine learning models are trained: XGBoost, CatBoost, Random Forest, Lasso, Voting Regressor, and others for comparative analysis.
• Model performance is evaluated using R², Mean Squared Error (MSE), Mean Absolute Error (MAE), and computational time.
• Hyperparameter tuning is applied to optimize model performance across all algorithms.
• The best-performing model is selected based on a composite evaluation of accuracy and efficiency.

Experimental results

Research questions

• RQ1 Which machine learning algorithm delivers the highest predictive accuracy for housing prices among XGBoost, CatBoost, Random Forest, Lasso, and Voting Regressor?
• RQ2 How do different models compare in terms of R², MSE, and MAE on the same housing price prediction task?
• RQ3 What is the computational efficiency of each model, and how does it impact real-time applicability?
• RQ4 Can a hybrid or ensemble model outperform individual models in housing price prediction?
• RQ5 To what extent can machine learning models support informed decision-making for homebuyers, sellers, and real estate brokers?

Key findings

• XGBoost achieved the highest R² score, indicating the strongest model fit to the housing price data.
• XGBoost recorded the lowest Mean Squared Error (MSE) and Mean Absolute Error (MAE), confirming superior prediction accuracy.
• XGBoost demonstrated the fastest computational time among all evaluated models, enhancing its practical deployability.
• CatBoost and Random Forest performed competitively but were outperformed by XGBoost in all key metrics.
• Lasso regression showed moderate performance but was less accurate than tree-based ensemble models.
• The Voting Regressor, combining multiple models, improved performance slightly but did not surpass XGBoost individually.