PERFORMANCE METRICS TREND ANALYSIS OF FEATURES PRESENT IN TRANSPORTATION SYSTEMS

摘要

Transit agencies frequently face the problem of having massive amount of performance data that await to be analyzed in order to determine what it is working and what it is not. It is beneficial to have automatic trend analysis methods that classify performance metrics of different scopes (elements of the transit system, e.g., stop, route, segment, whose performance is measured) into different categories. This is also important since agencies can analyze the effect of a change on a route by asking the algorithm to provide an assessment before and after the implementation of the change. In this paper we propose to analyze historical trends so that the outcome of this analysis is to group the routes that are performing well, the ones that do not experience any changes, the ones that are performing poorly, etc. The proposed method can easily be run in a parallel architecture when the amount of data to be processed is huge.