[ ] SIENA

Stimulation Initiative for European Neural Applications

Esprit Project 9811

Case Studies of Successful Applications

Predicting Sales of Articles in Supermarkets

Company background

The following application has been developed and tested in co-operation with a supermarket chain with a great number of stores in Fulda, Germany.

The problem

Time series prediction for economic processes is a topic of increasing interest. In order to reduce stock-keeping cost, an appropriate forecast of the future demand is necessary. A neural network research group of the university of Osnabrück was concerned with the grasping of the sales of articles in a supermarket and with the prediction of future demands.

Neural network application

Multilayer perceptrons have been trained on the known sale volume of the past for a certain group of related products. Additional information like changing prices and advertising campaigns have also been given to the net to improve the prediction quality. In this case, the information of 53 articles has been used. The information about the number of sold articles and the sales revenues in DM are given weekly. In addition, there were advertising campaigns for articles often combined with temporary price reduction. Such campaigns have had a significant influence on the demand for this article.

Benefits

It has been shown that neural networks can be trained to approximate the time series of sales in supermarkets and give an appropriate forecast on the future demand for articles.

Generalization

This application shows that neural networks can be trained to forecast the future demand for a special group of products on the basis of the past data in many economic processes even together with external information such as changing prices or advertising campaigns.

Contact Person

Frank M. Thiesing - Universität Osnabrück FB Mathematik / Informatik, D-49069 Osnabrück. Tel.: +49 541 969 2558, Fax: +49 541 969 2770.

Reference for further reading

Vemuri, V.R., Rogers, R.D.: Artificial Neural Networks - Forecasting Time Series. IEEE Computer Society Press 5120-05, 1994.