FORECASTING AUTOMOBILE DEMAND AND SALES IN THE NIGERIAN MARKET: A MACHINE LEARNING APPROACH TO URBAN MOBILITY, MARKET COMPETITION, AND POLICY INSIGHTS

Abstract

The estimation of automobile demand is central to both academic inquiry and policy planning, particularly given the sector‘s critical role in global economic activity. In developed economies such as the United States, Germany, and China, the auto industry serves as a paradigmatic case for analyzing market dynamics in differentiated, oligopolistic settings. Accurate demand forecasting is essential for production planning, pricing strategy, and infrastructure development. However, in emerging markets like Nigeria, empirical research on automobile demand remains sparse despite its growing relevance. Nigeria's automotive landscape is undergoing rapid transformation, propelled by urbanization, a rising middle class, and industrial policy reforms such as the National Automotive Industry Development Plan (NAIDP). This study addresses the empirical gap byevaluating the performance of various regression models, including the OLS, MARS, Regression Tree, Random Forest, and Gradient Boosting, in predicting automobile demand using real-world data. Among the models tested, OLS emerged as the most effective, with the lowest error metrics (MAE = 0.15, MSE = 0.06, RMSE = 0.24) and a strong explanatory power (R² = 0.86). In contrast, the MARS model underperformed, displaying the highest error rates and limited predictive capacity (R² = 0.43). Ensemble methods (RF and GB) showed moderate performance, with GB slightly outperforming RF in terms of relative error (MAPE = 0.01). The RegressionTree modelalsoperformed well,balancingaccuracyand interpretability.Thefindingsoffer valuable insights for both policymakers and industry stakeholders in Nigeria, emphasizing the importance of model selection in automotive demand estimation and the strategic implications for infrastructure and investment planning.