Track: Operations Management and Operations Research

Abstract

An objective visibility analysis has been developed as an integral part of the research  "Identifying Indicators for Rice Crisis Forecasting in Asia". Analysis of visibility is a topic that received less attention in the study of forecasting models. Yet without a visibility analysis predicted results would be questionable whether the forecasting results can still be trusted. Analysis of visibility is required for assessing the ability of a model to forecast future events. In this paper we introduce visibility error which has different concept from the definition of error obtained during model building. Visibility error is the forecasting performance statistic which is needed as base in determining how long period of forecasting can be done. Without this statistics the forecaster can predict the next thousand years without clear statistical reasoning. Specifically this study developed error time variant visibility using inductive logic. One important finding raised in this paper is a formula for determining the numbers of data testing required to perform forecasting work based on specific forecasting specifications. The formula is: N = V + 2.

Keywords: Visibility, error visibility, forecasting, error time variant