Classification of travelers for the recommendation of suitable integrated multimodal services

Abstract

This paper presents limitations of existing congestion management techniques which are attributed to the absence of robustness, coordination among traffic control elements and lack of adaptation in the current method to the changing traffic circumstance. The aim is to use clustering analysis in the area of machine learning to identify and develop classifiers that possibly “classifies” the types of transportation users, including travelers driving a car, riding a bicycle, riding a bus, taking a taxi and hiring a private car; which involves grouping the traveler as a driver or a passenger. For this study, statistical and analytical methods of distinguishing variations in the preferences of mode choice among the commuters was applied on a secondary data feedback collected from Birmingham city council. Euclidean distance is a clustering algorithm that was considered as attribute model, which includes distance algorithms i.e. distance between points and distributions models. Nine (9) questions and answers from the feedback were selected and modified for the statistical clusters modeling and the K-means algorithms was used in the clustering process. The result from the analysis was exported out in excel format in which the data of travelers was categorized mainly into two groups (Drivers/Passenger), which was further used to classify individual traveling behaviours. Certain rules were set identify types of Drivers and Passengers, persuadable and non-persuadable. These categories were further used to classify individual traveling behaviors. Certain rules were set to identify types of Drivers and Passengers, persuadable and non-persuadable.