9. Conclusions

While the results of these two case studies were contradictory, it appears that there might be some limited application of temporal models. It seems possible that there are cases where temporal models may offer advantages in predictive power over traditional all-time period models. However, it is also clear that there are significant barriers to fully adopting temporal modelling in place of traditional procedures.

Similarly, there are several issues which were not examined in the scope of this research which need some resolution in order to make temporal modelling an effective approach. Clearly the greatest drawback for temporal predictive modelling is that it creates a great deal more work to gather and analyse data, create and evaluate the model, when compared with a traditional predictive model. Without resolving these issues, it means that the added workloads may not justify the small gain in predictive power.

There are several unresolved questions, however, which should be further evaluated. The first of these is the management of the increased number of predictive models created when employing a temporal modelling approach. As mentioned, most clients of the CRM industry want definitive answers, rather than to have a series of models which need be consulted. However, if the solution was a combination of the models in some manner, this approach would cause a dilution of the temporal patterns which were shown in the course of this research.

A second area of potential research would be to examine how we can identify those places in which temporal predictive models will be effective and practical. This need is clearly demonstrated by the Winnipeg and MbMF study areas.

Finally the role of functional differences (i.e. locational differences between habitation sites, kill sites, and natural resource procurement sites) should also be investigated, with the goal of integrating functional and temporal differences into an overall model. However, it must be said that achieving this level of sophistication will come at the cost of high levels of repetition in the modelling process. For example, a project with three functional divisions and four temporal divisions will require the creation of twelve models! This approach would further compound the problems of how to integrate the separate models into a client-friendly presentation model.