Improving Inventory Performance with Clustering based Demand Forecasts

Proper management and better control of inventory of food items are one of the most important and important goals of food store managers. In this study, we try to provide knowledge of customer segmentation based on various characteristics as inputs in predicting retail demand. The purpose of this paper is to provide a prediction model for retailers based on customer clustering to improve inventory performance. Customer clustering with the genetic algorithm is performed in MATLAB R2016a software. The proposed model is used to predict the demand for five items of supermarket goods in Gorgan. In this paper, the predictions of ARIMA, ARIMA, MLP, and GMDH neural networks are used to predict. Modeling of these models has been done in MATLAB software. The results showed that the GMDH neural network with the clustering of customers had the least predictive error. The model predicted by the inventory control policy reduces the number of days that the shortage faces and increases the level of service to the customer. Retailers can use the proposed model to predict the demand for various items to improve inventory performance and profitability of operations.