Machine Learning Insights into Retail Sales Prediction: A Comparative Analysis of Algorithms

The knowledge of machine learning is expanding across all parts of existence, and it is an influential force of implemented functions. They have been implemented in education system, healthcare sector, engineering, retail sales and many other fields and its application is increasing. As the field of retail continuously grows and more companies enter this field, data processing and machine learning play crucial roles in determining or estimating the sales. The conventional approaches of sales that do not rely on comprehensive information are not efficient in the current market. Due to the growth of machine learning, various factors including but not limited to purchase behaviour, target market, and future sales have been determined making planning and growth easy. This paper aims at providing a detailed description of the application of ML methods to make predictions of retail sales with respect to linear regression, random forest and XGBoost models. The purpose is to determine which of them can be used by the retailers for decision making and which contributes to higher predictive value. All the models employed were trained and tested using the Big Mart sales recorded data that is publicly available. When using various regression models, the highest R-squared values of 0.545 were estimated by Random Forest Regression. Thus, this research aims to advance the usefulness of sales forecasting by applying and comparing the outputs of these models to real life retail data.