Demand Forecasting with Clustering and Artificial Neural Networks Methods: An Application for Stock Keeping Units

Introduction: Firms’ production strategy consists of interrelated strategic decisions, including pricing, demand forecasting and demand response planning, capacity planning, capital and cost structure. Production strategy is also the most important factor affecting the overall return of companies. Companies must estimate future situations in order to maintain their current position. Nowadays, with the developing technology, companies contain a wide variety of data. With various data analytics and optimization methods and tools, the data that can help the decision making process of companies can be made meaningful and usable. Objective: For companies like sanitary wares with large number of product variants, product groups based on estimated requirements will emerge. The ceramic sanitary ware sector, where the product variety is very high, shows seasonal effects in itself and seasonal and trend effects in its products. This situation makes it difficult to make estimations in the ceramic sector. The aim of this study is to increase the accuracy rate of demand estimation ratio by more than 70%. Methods: First of all, k-means clustering algorithm is used to obtain product groups. Then, an artificial Neural Networks model is used to estimate demands of product groups. Results: The obtained estimation error ratios are compared with those which are obtained by exponential smoothing and moving average methods in time series methods. It is observed that the most suitable method is Artificial Neural Networks (ANNs) for obtaining best results. Conclusion: The results prove the efficiency of the applying ANNs to clustered products as a nature-inspired method for demand estimation problem.