Statistical and genetic algorithms classification of highways

This paper reports the results of experiments comparing a conventional stat istical method and an evolutionary genetic algorithms approach for classify ing highway sections that is based on temporal traffic patterns. Traffic pa tterns are used as surrogates of two important characteristics of a highway section, namely, trip purpose and trip length distribution. Accurate class ification can lead to better traffic analyses, such as estimations of annua l average daily traffic volume and design hourly traffic volume, and determ ination of maintenance and upgrading schedules. Modern-day computers cannot solve the problem of obtaining optimal classification corresponding to min imum within-group error. The hierarchical grouping method provides a reason able approximation of the optimal solution. However, for smaller numbers of groups, the hierarchical approach tends to move farther away from the opti mal solution. The genetic algorithms based approach provides better results when the number of groups is relatively small (e.g., less than nine for th e Alberta highway network). In addition to comparing the two methods, the r esults of additional experiments studying the characteristics of the geneti c approach are included.