Transit system monitoring and design
Yorgos J Stephanedes
Report no. Mn/DOT 1994-17
Statistical techniques were developed for extracting the most significant features (indicators) from a transit system data base, and classifying proposed and existing transit systems according to the selected features. The data base was constructed by using information from all previous years available by the Mn/DOT, the Census and other sources to be used in classifying transit systems. The data base emphasized the use of raw characteristics of the operating system and the area socioeconomics. The feature extraction was done so that the minimum number of features were extracted that can be used for classifying the transit systems with maximum accuracy. The classification method was designed around the data base and is flexible so that it can use future data to update the data base at minimum cost. The transit system patterns, resulting from the classification method, were identified according to need and performance, and the main characteristics were specified for each pattern. These characteristics and descriptions identifying each pattern determines whether it should be modified. A controlled experiment was required to test the classification method. A randomly selected part of the data was classified by the method, and then the unselected data was treated as a control group for the experiment. After the experiment a percent of misclassifications was calculated.