In recent years, several hybrid choice models were developed to incorporate the psychological factors into the discrete choice models that have been widely applied in transportation. The psychological factors refer to the individuals' attitudes, which are usually measured in the form of interval scale such as 5-point scale (very poor, poor, indifferent, good and very good). These linguistic variables are first transformed to numerical data before the mode estimation. Up to now, three representative approaches have been widely applied to incorporate these linguistic variables (psychological factors) into the conventional models. The first approach introduces these psychological factors directly into the utility function. In contrast, the second one first calculates the latent variables from factor analysis on these psychological factors and then incorporates them into the model. However, it is difficult to precisely distinguish the boundaries of the linguistic variables. For instance, even the linguistic value is "poor", it might mean "very poor" at the possibility of 40% and "poor" at 80%. Therefore, it is necessary to represent the fuzziness of the linguistic variables properly in the model to make a full use of data information. In this paper, fuzzy inference using fuzzy number and approximate reasoning is employed to represent the fuzziness of the linguistic variables. The aim of this paper is to suggest a new type of hybrid choice model that incorporates individuals' attitude into discrete choice model based on fuzzy inference systems (FIS), regarding individuals' attitudes about the corresponding alternatives. Since the introduction of FIS does impose any extra restriction on the error terms of utility function, the resultant hybrid model can be estimated based on maximum likelihood estimation method. To confirm the effectiveness of the proposed hybrid model, an SP (Stated Preference) survey data was used. This data was collected in 1999 to evaluate the effects of real time traffic information on the switching behaviour from private car to a light rail system, called 'Amstramline', in Hiroshima city, Japan. In the survey, the respondents rated their satisfaction degree for the following attributes of private car, access bus, and Amstramline: (1) the traffic information of private car (travel time, congestion, radio, and television), (2) the level-of-service of access bus (punctuality in ordinary day, punctuality in rainy or snowy day, frequency, travel time, and fare), and (3) the level-of-service of Amstramline (punctuality in rainy or snowy day, frequency, operation time, travel time, and fare). The FIS is applied to infer the satisfaction degree for each travel mode. The effectiveness of the suggested hybrid model is examined based on the comparison with the conventional model. The estimation results show that most of the psychological factors are statistically influential in the mode choice. Especially, the satisfaction degree of the access bus is the most important psychological factor in the mode choice. The hybrid models were found to be superior to the conventional logit model. For the covering abstract see ITRD E126595.
Abstract