Sales forecasting for a Turkish paint producer: Artificial intelligence based methods versus Multiple Linear Regression

Sales forecasting has a great impact on facilitating effective and efficient allocation of scarce resources. However, how to best model and forecast sales has been a long-standing issue. There is no best method of forecasting in all circumstances. Therefore, confidence in the accuracy of sales forecasts is derived by corroborating the results using two or more methods. This paper evaluates the relative performance of Linear Multiple Regression, Artificial Neural Networks and Adaptive Neuro Fuzzy Networks by applying them to the problem of sales forecasting for a Turkish paint producer firm. The results indicate that Adaptive Neuro Fuzzy Networks yields better forecasting accuracy in terms of Root Mean Square Error and Mean Absolute Deviation.