Modeling of Consumer Product Sales using an Adaptive Neuro-Fuzzy Inference System

Sales modeling is essential for businesses in today’s strongly competitive world. The sales forecasting of products under different retail format has become a major concern of companies for evaluating the circulation and competitiveness of their products. Due to the presence of a wide range of variable product characteristics, it is very complicated to model the sales of consumer products. The present study aims to design an adaptive neuro-fuzzy inference system (ANFIS) to model the sales of consumer products under different retail format. Sales were investigated based on the views of consumer product experts through the ANFIS technique. As a new technique for sales estimation, ANFIS is a combination of neural network learning and fuzzy logic generalization. The input and output criteria were determined by performing in-depth interviews with twelve sales experts and managers and collecting the average sales data of 336 products in 2019. The results indicated that ANFIS-based sales modeling could help manufacturers of consumer products more accurately forecast retail type demands. Finally, the training and testing efficiency of the proposed model was evaluated using the root mean square error (RMSE), mean absolute error (MAE), standard error of the mean (MSE), and relative absolute error (RAE).