METHODOLOGY FOR ANALYZING ERRORS IN PREDICTION WITH DISAGGREGATE CHOICE MODELS

Predictions of future travel behavior and of the performance of alternative transportation systems are needed by transportation planners and decision makers to determine the desirability of alternative transportation plans. The usefulness of predictions, and consequently of prediction methods, depends on their accuracy. This paper presents a methodology for analyzing errors in prediction with disaggregate choice models. The paper describes the process by which disaggregate choice models are formulated and used for prediction. The sources of error in the model formulation and prediction process are identified. The interaction and propagation of these errors to the final prediction are analyzed. A set of error measures is proposed for evaluating the performance of alternative prediction models. A strategy is developed for analyzing the source of different components of the total error. An empirical analysis of errors in the prediction of mode choice to work illustrates the use of this approach for evaluating the accuracy of a set of prediction models, identifying major sources of error in prediction, and suggesting steps that can be taken to improve these prediction models. /Author/

Media Info

  • Media Type: Print
  • Features: Figures; References; Tables;
  • Pagination: pp 17-23
  • Monograph Title: Perception and values in travel demand
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00145325
  • Record Type: Publication
  • ISBN: 0309025613
  • Files: TRIS, TRB
  • Created Date: Feb 16 1977 12:00AM