Event Details

A new classification scheme for trend/seasonal demand forecasting and planning applications

When dealing with large data sets of historical consumer demand data, it becomes important to classify trend and seasonal patterns into manageable groups for automated forecast modeling. Demand planners in consumer driven industries encounter sales data with varying degrees of trend and seasonality. However, the critical tools necessary for demand planners to effectively analyze and forecast such demand data are primarily based on best single model performance criteria.

In this talk, I examine the relationship in the strengths of trend and seasonal variation in historical time series., I introduce a new, readily implementable, algorithm for classifying large datasets into six non-overlapping groups of trend/seasonality strengths in the sales data and illustrate this with a number of data sets across multiple industries.