OPTIMISING NIGERIAN RETAIL DEMAND FORECASTING: LEVERAGING MACHINE LEARNING AND EXTERNAL FACTORS

Abstract

Nigeria’s dynamic retail sector presents complex forecasting challenges due to market volatility and socio-economic factors. This study develops a machine learning-based solution using historical sales data and external variables (weather, fuel prices, CPI, unemployment) sourced from Kaggle. Linear Regression, Decision Trees, Random Forests, and Feedforward Neural Networks were evaluated using MSE, MAE, MAPE, and R². Feature engineering revealed price (26.7%) and promotions (17.3%) as dominant demand drivers. The Random Forest model outperformed others, reducing MAPE from 25.4% to 16.6% and improving R² from 0.216 to 0.400 when incorporating external factors. These results highlight the value of integrating socio-economic indicators into forecasting systems, offering Nigerian retailers a scalable, Machine Learning-driven framework for inventory optimization. Future research should explore hybrid models and real-time signals, such as social media trends, to enhance predictive accuracy.