Hedonic Price Determinants and Regression Benchmarking Forresidential Property Valuation in Nigeria

This study examines determinants of residential property prices in the Nigerian urban housing market and benchmarks three linear regression models using 13,932 residential transactions. Pearson correlation analysis identifies the principal structural and spatial predictors, while Ordinary Least Squares (OLS), Ridge and Lasso regression are evaluated using a CRISP-DM-based workflow incorporating log transformation, outlier diagnostics, multicollinearity assessment and 10-fold cross-validation. Total living area shows the strongest positive correlation with log sale price (r = 0.713), followed by special features value (r = 0.506) and land area (r = 0.370); subcenter distance has a negative association (r = −0.404). OLS achieves the highest reported performance (RMSE = 0.5097; R² = 0.778), while Lasso and Ridge yield R² values of 0.778 and 0.774, respectively. All thirteen predictors are statistically significant in the OLS specification, with living area (t = 89.67), structural quality (t = 50.12) and property age (t = −31.91) showing particularly strong associations. The maximum variance inflation factor is 6.94, indicating no severe multicollinearity under the stated threshold. These findings provide an interpretable linear modelling benchmark for the dataset and underscore the importance of combining structural attributes with multiple spatial accessibility measures in residential valuation.