Further Topics

Abstract

The information provided in Chapters 1 through 10 should be sufficient to provide an understanding of the processes involved in the construction and application of predictive models of consumer behaviour. In this chapter three further topics are discussed. The first considers model construction and validation when only small samples are available. The second topic looks at alternative methods of assessing model performance, which are particularly useful when two competing models are being compared. The third and final topic examines multi-model systems, sometimes referred to as ensembles or fusion systems. A multi-model system is when several different models are constructed to predict a single behaviour. The predictions made by each model are then combined together in some way to deliver a better overall prediction than that generated by any single model on its own.