CART Ensemble and Bagging Algorithm for Estimating of Factors Influencing the Furniture Market

Abstract

Demand and sales forecasting are critical to making successful marketing decisions for any business in all industries. This study investigates the influence of various factors on the turnover of a large furniture company in Bulgaria. Factors include customer flow and the company's advertising strategies – TV and Internet (on FB and Google). The customer flow is presented separately for visitors to company stores and on the company's website Daily observations for nearly two years are modeled using the powerful machine learning technique CART Ensemble and Bagging. All constructed models are trained using the cross-validation to obtain quality predictions The built models describe furniture sales with high goodness-of-fit statistics: coefficient of determination 96%, MAPE under 5% and determine the factors affecting sales. The models are applied to forecast the turnover for seven days ahead.