Abstract
We present a pattern analysis on the booking curve of an intercity passenger railway. It consists of three steps: cluster analysis, discriminant analysis, and prediction analysis. Basic booking curve patterns are built on the base of risk preference and booking magnitude. Using four-week ticket sales data from Taiwan Railway Administration, a numerical case study of the rapid train on a long distance market is illustrated to demonstrate the characteristics of the proposed method. We obtain pretty good statistical results in cluster analysis and discriminant analysis. Although its overall prediction accuracy is not good, the proposed method provides a useful pattern analysis procedure. Many forecasting techniques of growth curve may be included in the process so as to improve overall fit of the method in the future.
