European Transport / Trasporti Europei (2007) 37/XIII

In the urban areas, public transport service interacts with the private mobility. Moreover, on each link of the urban public transport network, the bus speed is affected by a high variability over time. It depends on the congestion level and the presence of bus way or no. The scheduling reliability of the public transport service is crucial to increase attractiveness against private car use. A comparison between a Radial Basis Function network (RBF) and Multi layer Perceptron (MLP) was carried out to estimate the average speed, analysing the dynamic bus location data achieved by an AVMS (Automatic Vehicle Monitoring System). Collected data concern bus location, geometrical parameters and traffic conditions. Public Transport Company of Palermo provided these data.

This paper applies a multinomial logit model to the choice of a departure railway station by Dutch railway passengers. This is a relevant theme since about 50% of Dutch railway passengers do not travel via the nearest railway station. The passengers’ choices for departure stations are aggregated at the four digit postal code area level. We applied three functional forms for the underlying systematic utility of a station, namely a linear effect of attributes, cross effect of distance and frequency of service, and a translog formulation on distance and frequency of train services. With 3,498 post code areas and 360 railway stations our analysis found consistent effect sizes for distance, frequency of service, intercity status of the station and the presence of park-and-ride facility on the choice of departure station. The effect of distance on the choice of a departure station declines smoothly. The effect of frequency of service is relatively small compared to the effect of distance. A frequency of service increase by a hundred trains per day is equivalent to being 600 m closer to the station. The Intercity status of the station plays the biggest role in the choice of departure station. It has an equivalent effect of a change in 2 km distance or about a frequency of service of 300 trains per day. In addition, the presence of park-and-ride facility in the station poses a sizable effect in the departure station choice. In most cases its effect reaches about 35% of the intercity status effect.

Traffic violations are among the leading causes of road accidents. In this research, the sensitivity of Greek drivers to a hypothetical intensification of police enforcement for speed violations and improper overtaking is analyzed, using stated preference data. Under the assumption of increased police enforcement, drivers were presented with the option to maintain their unsafe driving patterns (and risk getting fined) or comply with the traffic laws (and experience longer trip duration). A parsimonious mixed logit model has been estimated and sensitivity analysis is performed with respect to the main variables. The model explicitly captures the (unobserved) heterogeneity in the sample, and reflects the fixed random parameter across observations from the same respondent. The behaviour of the surveyed drivers depends on socioeconomic characteristics and trip characteristics. Based on the presented sensitivity analysis, it can be argued that while the “typical” Greek driver may not be particularly riskprone, there are segments of the population that show a tendency to violate traffic laws. This is a useful finding that could be used by policy makers e.g. to develop targeted police enforcement campaigns (or targeted media campaigns, special education initiatives, etc.), aimed at the demographic segments with a higher tendency for traffic violation.

Modellers are increasingly relying on the use of continuous random coefficients models, such as Mixed Logit, for the representation of variations in tastes across individuals. In this paper, we provide an indepth comparison of the performance of the Mixed Logit model with that of its far less commonly used discrete mixture counterpart, making use of a combination of real and simulated datasets. The results not only show significant computational advantages for the discrete mixture approach, but also highlight greater flexibility, and show that, across a host of scenarios, the discrete mixture models are able to offer comparable or indeed superior model performance.