doi: 10.18178/JAAI.2025.3.1.1-18
Let’s Boost House Price Predictions: A Machine Learning Approach for Norwich
J. D. Adekunle
1,*
,
,
,
,
,
,
- 1 Department of Mathematic, Federal University of Agriculture, Abeokuta.
- 2 Department of Statistics, Federal University of Agriculture, Abeokuta.
- 3 Graduate school of Asia Pacific studies Ritsumeika, Asia Pacific University.
- 4 Medipolis GmbH, Otto-Schott-Straße, Jena, Germany.
- 5 Department of Management Information Systems, Topdel Engineering Limited, Lagos, Nigeria.
- Manuscript received November 20, 2024
- revised November 29, 2024
- accepted December 12, 2024
- published January 17, 2025
Abstract
In recent years, the demand for accurate housing price predictions has intensified, driven by the dynamic nature of real estate markets and the need for data-driven decision-making. Machine learning models (a subset of AI) have emerged as powerful tools in this domain, offering enhanced predictive capabilities over traditional statistical methods. In this paper, we aimed to predict house price in Norwich and evaluate the factors that drive the price. To achieve this, we trained four boosting (Gradient Boosting, XGBoost, LightGBM, and CatBoost) to predict the house price. The performance of these models was evaluated in a standard evaluation approach and post-hoc residual evaluation approach within three designed instances (testing, training, and combined [testing + training]). The predictive performance and significant predictors were identified, with Beds, Baths, Sqm, and other features showing high significance, while age of the house was not significant. We found out that GradientBoost and XGboost are closely related in their residuals, while LightBoost operates independently. The performance metrics revealed that LightGBM outperformed the other models with the lowest Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) in both training (RMSE [5.891], MAE [3.680]) and test (RMSE [13.170], MAE [7.092]) instances, achieving an R-squared value of (combined [0.99] train [0.998], and test [0.99]). Correlation analyses of the residuals indicated a strong positive correlation between Gradient Boosting and XGBoost (train [0.84], test [0.85], combined [0.84]), while CatBoost demonstrated a moderate correlation with both. Notably, LightGBM (−0.04 ≤ r ≤ 0.3) exhibited distinct residual patterns, showing no significant correlation with the other models, suggesting it captures different aspects of the dataset. These findings show the importance of utilizing an ensemble approach that includes LightGBM to enhance predictive accuracy by leveraging its unique error characteristics alongside the complementary strengths of the other models, and inform model selection and ensemble strategies in future.
Keywords
- Boosting algorithm
- house price
- real estate valuation
- residential property prices
- norwich housing market
How to Cite
J. D. Adekunle, M. I. Oyeniran, H. S. Sule, T. T. Akinpelu, E. J. Ayanlowo, C. K. Ogu, and C. O. Robert, "Let’s Boost House Price Predictions: A Machine Learning Approach for Norwich," Journal of Advances in Artificial Intelligence, vol. 3, no. 1, pp. 1-18, 2025. https://doi.org/ 10.18178/JAAI.2025.3.1.1-18
Copyright & License
Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).