Sales prediction in the ice category applying fuzzy sets theory

Abstract

With growing pressure on performance and data regarding customer behaviour and supply chain process widely available, stock keeping units aim to optimise the level of inventories. It is natural that good estimates of future sales can substantially increase the efficiency of the overall company. We can distinguish two basic perspectives: one assumes sales to be an independent process; the other explores its dependency on exogenous variables. In this paper we focus on the forecasting of sales in the Ice category when dependency on quarterly average temperatures in the form of exponential function is assumed. We concentrate especially on LFL-Forecaster, a method combining fuzzy transform and fuzzy natural logic of fuzzy sets theory, as a tool for average temperature forecasting. The results are compared with simple linear extrapolation and truly observed temperatures. The utilisation of LFL-Forecaster is found to be superior to simplifying linear regression.