Ensemble Strategies for Predicting Walmart Sales Using Regression Models

This study looks into predictive modeling in the retail sector, with a specific focus on improving sales price predictions for Walmart’s extensive product line. It uses advanced machine learning algorithms and statistical models to rigorously analyze historical sales data alongside external factors such as economic indicators and seasonal trends in order to generate effective models. Our research uses approaches such as Light GBM, Random Forest, Gradient Boosting, Cat Boosting, and K Nearest Neighbors to explore the complex interplay between pricing dynamics and consumer behavior. Our findings not only promote predictive analytics in retail, but also have practical implications for industry practitioners, particularly Walmart, in terms of pricing strategies, inventory management, and overall business performance. We show that the voting ensemble strategy outperforms averaging by accurately improving predictive accuracy across varied base model combinations.