Modeling Residential Property Prices in Emerging Climate-Responsive Urban Markets: A Hybrid Modeling Framework for Baidoa City-Somalia

Mohamed Ibrahim Nor^1,2*Shuaib Nour Hussein^1,3

• ¹SIMAD University, Mogadishu, Somalia
• ²Institute of Climate and Environment (ICE), Mogadishu, Somalia
• ³Graduate School, Mogadishu, Somalia

This study aims to examine the determinants of residential property prices in Baidoa's climate-responsive real estate market. It investigates both linear and non-linear interactions among key variables to enhance property valuation models and inform urban development strategies. A hybrid-methods design was adopted, integrating a hedonic regression model with an artificial neural network (ANN) framework. The analysis utilizes a stratified random sample of 118 residential properties from the Baidoa Housing Survey, capturing diverse features such as property size, number of bedrooms, proximity to the central business district (CBD), safety, age, and air quality. Model performance was evaluated using standard metrics (e.g., R², MSE, MAE) along with diagnostic tests and 5-fold cross-validation. The hedonic regression model explained 74.2% of the variance in property prices, with key variables showing significant influences. The ANN model further reduced prediction errors by approximately 20%, effectively capturing complex nonlinear relationships among the predictors. Compared to the baseline linear hedonic regression model, the ANN achieved approximately a 20% reduction in mean squared error (MSE), with performance improvements validated through 5-fold cross-validation and supported by 95% confidence intervals. The results underscore the importance of strategic urban planning interventions such as improving neighborhood safety, enhancing infrastructure near the CBD, and boosting environmental quality. These insights offer practical guidelines for policymakers and real estate practitioners to foster sustainable urban growth and more accurate property valuations. This research uniquely combines traditional econometric methods with advanced machine learning techniques, yielding a hybrid model that outperforms conventional approaches. Its application to Baidoa-a rapidly urbanizing city facing distinct socioenvironmental challenges-adds novel perspectives to climate-responsive real estate market analyses. The study bridges the gap between classic hedonic pricing theory and contemporary neural network methodologies, providing both theoretical and empirical contributions. It extends current understanding of urban housing market dynamics and offers a robust analytical framework that can be adapted to similar emerging urban contexts.