Reducing the aggregation bias: A heuristic approach

Abstract

Disaggregate travel demand models are increasingly applied to predict aggregate demand. This is usually done by applying these models to zonal aggregated data. This prediction method causes a bias (“aggregation bias”) in the predictions obtained. Several aggregation procedures aimed at reducing this bias have been developed in recent years. Many of them, however, suffer from a number of drawbacks. First, no aggregation procedure can assure a priori the reduction of this bias to negligible proportions. Second, the direction of the residual bias is unknown. Third, no measure is provided for the residual aggregation bias. This paper presents a new approach to aggregate analysis. It is based on a heuristic model, which enables the application of diaggregate models to zonally aggregated systems for travel prediction. This heuristic decomposes the bias into two components: average systematic bias and a random variation around it. The systematic component is eliminated, such that the residual aggregation error is unbiased, its magnitude is small and a measure is provided to express this error quantitatively. The performance of the heuristic was demonstrated by an empirical study of mode choice in Washington, D.C. The heuristic was found efficient in reducing the bias and predicting the residual aggregation error.