Judgmental Time-Series Forecasting Using Domain Knowledge

Abstract

This chapter concerns principles regarding when and how to use judgment in time-series forecasting with domain knowledge. The evidence suggests that the reliability of domain knowledge is critical, and that judgment is essential when dealing with "soft" information. However judgment suffers from biases and inefficiencies when dealing with domain knowledge. We suggest two sets of principles for dealing with domain knowledge-when to use it and how to use it. Domain knowledge should be used when there is a large amount of relevant information, when experts are deemed to possess it, and when the experts do not appear to have predetermined agendas for the final forecast or the forecast setting process. ForecasterS should select only the most important causal information, adjust initial estimates boldly in the light of new domain knowledge, and use decomposition strategies to integrate domain knowledge into the forecast. Forecasters are often faced with sach questions as-Should we forecast judgmentally? Should we merely take the output of a statistical model? Should we adjust the statistical forecast for the knowledge we have of the environment that is not available to the models? The advantage of judgment over a pure statistical forecasting approach is that it can incorporate a great deal of domain knowledge into the forecasting process. Consider a typical sales forecasting meeting. Generally some past time-series history provides the basis of prediction. However, most of the discussion at such meetings centers on the contextual information that is relevant to the task (Lawrence, O'Connor and Edmundson 2000). This