Product Sales Forecast Model Considering Circular Fluctuations

Abstract

Retailers often make sales forecasts based on the purchase history data they have. When making forecasts, sales may be influenced by periodic factors such as fashion and climate, so the forecast model may consider periodicity. However, most of the proposed models deal with seasonal fluctuations. There is no doubt that seasonal fluctuations are important, but in order to improve prediction accuracy, it is necessary to pay attention to circular fluctuations as well. In this study, we identified the circular fluctuations in sales by Fourier analysis, and constructed a model with higher prediction accuracy than the conventional ARIMA model by taking that factor into consideration.