Analysis of transit quality of service through segmentation and classification tree techniques Oña López, Rocío de Oña López, Juan José De Bus transit Classification and regression trees (CART) Data mining Perceptions about the quality of service are very different among public transport (PT) users. Users’ perceptions are heterogeneous for many reasons: the qualitative aspects of PT service, users’ socio-economic characteristics, and the diversity of tastes and attitudes towards PT. By analysing different groups of users who share a common characteristic (e.g. socio-economic or travel behaviour), it is possible to homogenise user opinions about the quality of service. This paper studies quality as perceived by users of the metropolitan transit system of Granada (Spain) through a classification tree technique (classification and regression trees (CART)) based on five market segmentations (gender, age, frequency of use, reason for travelling, and type of ticket). CART is a non-parametric method that has a number of advantages compared to other methods that require a predefined underlying relationship between dependent and independent variables. The study is based on data gathered in several customer satisfaction surveys (non-research-oriented) conducted in the Granada metropolitan transit system. The models' outcomes show that some attributes are very important for almost all the market segments (punctuality and information), while others are not very relevant for any of the segments – most notably fare, despite the fact that fare was stated as very important by most of the passengers during the interview 2024-01-19T12:53:21Z 2024-01-19T12:53:21Z 2015 info:eu-repo/semantics/article Published version: Rocío de Oña & Juan de Oña (2015) Analysis of transit quality of service through segmentation and classification tree techniques. Transportmetrica A: Transport Science, 11(5), 365-387 https://hdl.handle.net/10481/86972 10.1080/23249935.2014.1003111 eng http://creativecommons.org/licenses/by-nc-nd/4.0/ info:eu-repo/semantics/openAccess Attribution-NonCommercial-NoDerivatives 4.0 Internacional