A data mining approach to forecast behavior

This study presents a data mining analysis of customer forecasting patterns of multiple customers (auto manufacturers) from a large auto parts supplier. We consider a manufacturing environment in which forecasts of future orders are used as inputs for a series of decisions. We define the complexities that are captured from our data set, developing the daily flow analysis to obtain accuracy ratios of forecasts as a performance measure for customers. We also demonstrate the application of some recent developments in clustering and pattern recognition analysis which can have a significant impact on the performance analysis of customers.