Walmart Sales Forecasting using Different Models

Abstract

This research study compares three regression models, namely Random Forest, Linear Regression, and Lasso Regression, to determine which model has better performance on predicting Walmart sales. Through the analysis of historical sales data, including factors such as time, unemployment rate, CPI, and temperature, the dataset have training and testing sets. Random Forest is implemented and compared with Linear Regression, a traditional statistical method, as well as Lasso Regression, which includes a regularization term for feature selection and prediction accuracy improvement. Performance evaluation is conducted using mean squared error,and R-squared score. The results consistently show that Random Forest outperforms both Linear Regression and Lasso Regression in predicting Walmart sales, demonstrating its accuracy and robustness. This research offers insights into predictive modeling in retail sales forecasting and highlights the potential for using Random Forest as a reliable tool for inventory control, demand forecasting, and strategic planning at Walmart and similar retailers. Overall, this study contributes to the understanding of sales prediction in the retail industry, suggesting avenues for future research in exploring advanced machine learning algorithms and data preprocessing techniques to further improve accuracy.