In this thesis, the functioning of passenger delay-based refund schemes in urban public transportation settings is evaluated, such as metros or light-rails. These are used when a passenger encounters a delay on their trip, in order to grant a form of refund to them. To the best of knowledge, no research has been carried out yet on several aspects of these refund schemes, such as which design considerations can be taken into account upon establishing one, or how their performance can be evaluated. To clarify these issues, the state of the practice is reviewed by surveying the internet and the information that is being provided by the individual public transport operators. By categorising this information, it can be established what the design aspects of these refund schemes are. These are the delay type such as departure or arrival delay, the delay threshold, i.e., the minimum delay from which a trip taken is elegible for a refund, the refund type, which is the form of refund that the passenger receives, and the possibility for alternative transport to be refunded. In order to determine whether there could be a form of linearity between delays and refunds, and to facilitate the evaluation of different refund schemes, a regression analysis was carried out on approximately a year of passenger trip data and vehicle data from the Washington Metro. This enabled acquiring more reliable, i.e., excluding individual passenger’s possibly erroneous behaviour, observations of the departure delay and arrival delay of every trip taken during this study period. Using this data, a possible link between the total amounts of delays encountered and refunds granted per day was analysed. It could be concluded that there is indeed a strong linear effect between these two. Furthermore, a number of key performance indicators (KPI’s) was established which allowed to assess the performance of a set of synthetic refund schemes, in which the parameters were varied that resulted from the aforementioned survey. Furthermore, a refund scheme was recreated that uses the general punctuality on a line and subsequently refunds all travelers that can have been expected to be effected by a possible drop in this punctuality. Using these KPI’s, conclusions could be drawn on the performance of different refund schemes, in the sense of how many refunds per trip are granted, how the ratio of the average delay of all refunded trips vs. the average delay of all trips is, how many revenue losses the refund schemes yield for the operator, and how much of the time losses (using a certain Value of Time) are compensated for.

Moving block signalling promises a significant reduction of the infrastructure occupation compared to a fixed block system, such as NS’54/ATB. This is mainly caused by a strong reduction of the approach and running time. However, to assess the capacity gains track occupation data, such as blocking times are needed. ETCS L3 Moving Block is still in development, so gathering data out of daily operations is not possible. Also gathering moving block data out of simulation isn’t always a convenient solution. For example, in simulation software FRISO, used at ProRail, ETCS L3 Moving Block is not (yet) implemented.
With infrastructure data, rolling stock parameters and planned time-distance data, blocking times for a moving block signalling system can be estimated. The model presented in this thesis has an average error of 0.87s to the blocking time. In 95% of the cases the error is within a range of (-3,3) seconds. Given that ProRail plans with a precision of 6 seconds, it can be concluded that the model both precise as accurate. With the blocking times of all trains, bottlenecks in both railway corridors and complex nodes can be identified. One could sum all blocking times per block and consider blocks with the highest summed blocking times as bottleneck. However, this can only be applied for homogeneous traffic situations. Another approach is identifying bottlenecks by the shortest buffer time between two trains, also called a critical block. This can be applied for both homogeneous as heterogeneous traffic situations. An advantage is that it is not needed to split corridors into line sections. One could analyse a whole network at once and identify bottleneck at a microscopic level. By keeping the same timetable, the buffer time between two trains increases on average by 75 seconds (60%) using moving block over NS’54.

The provision of good accessibility is the main function of a well-functioning public transport (PT) system, the quality of which affects the ability of people to access basic needs such as employment, education, and healthcare. It is therefore important that PT accessibility is distributed in an equitable, or fair, manner. This thesis evaluates the equity of a PT network according to three distinct distribution principles, defined as alternative ideas of what constitutes a fair distribution of resources. Egalitarianism, proportionality, and sufficientarianism are applied in separate equity evaluations of the same area, resulting in the identification of zones of surplus and deficit accessibility according to each principle. When comparing the results of the three equity evaluations, it is found that both the locations and magnitudes of accessibility surpluses and deficits differ between principles. Some principles have deficit zones in common at the periphery of the study area, indicating that the borders of concessions may have been somewhat neglected in the service planning process. However, overall, the main finding is that a PT network that is equitable according to one principle may not be equitable according to another. It is also attempted to use the results of each equity evaluation to make frequency modifications to achieve a more equitable PT network according to each principle. Equity evaluation according to egalitarianism and proportionality were found to not be possible due to the circular calculations inherent in the methodology. It was possible to use the results of the sufficientarianism equity evaluation in network design, therefore this principle can be used as an input in the network planning process. Although further research is still required, these findings can help to inform the incorporation of equity in future PT planning policy, for example through inclusion in concession requirements.

Inaccurate truck cycle time (TCT) prediction in earthworks impacts construction projects because more equipment and human resources must be added to complete the project. It also increases fuel consumption and emissions from the machinery. However, the current method gives inaccurate prediction because of subjectivity and human error.
This research aims to utilize the historical data for improving the accuracy of TCT prediction in earthworks. Two historical data are explored, such as manual and automated data provided by BAM, using a machine learning approach in which regression techniques: Multi Linear Regression (MLR), Support Vector Regression (SVR), and Artificial Neural Network (ANN).
The result concluded that automated data could develop predictive models because of its quality and variance. ANN develops most predictive models with feature combination one or two, where is distance is the important feature. The benefits of the predictive model are analyzed by comparing them with the traditional method in predicting truck cycle time. The models are more accurate in predicting truck productivity and have lower inefficient truck cycle time than the traditional method. The reduction of inefficient truck cycle time can reduce the cost for fuel and drivers, fuel emission. Models also have intangible benefits to stakeholders, such as gaining partners trust and a better strategic plan to complete the project.
In conclusion, the historical data can improve the prediction accuracy of TCT and give benefits to stakeholders. And practitioners are suggested to raise awareness about the importance of data and improve earthmoving documentation for better predictive models.

As traffic demands are ever increasing and building new infrastructure poses challenges in densely populated areas, it is important to optimally utilise existing infrastructure. Short-term traffic forecasting can help with this task, as its predictions can help to prevent congestion by rerouting vehicles. Recently, neural networks developed for traffic forecasting have lead to an unprecedented accuracy, but deploying these on large scale can be difficult as the resulting models likely overfit to the highway they were trained on. This thesis therefore performs an in-depth assessment of the accuracy of neural networks traffic speed predictions on highway stretches containing different types of congestion patterns. By relating the results back to traffic flow theory, these results can be put into perspective.

This research explored two types of models, ARIMA and multiple linear regression, for forecasting tourist counts in 7 locations around Amsterdam red light district, for the prediction horizon of up to 30 minutes.

To address the increasing passenger demand in the coming years and make public transport less crowded and delayed, insights into predicted passenger flow are needed. A wide range of studies has used and validated that smart card data can be one of the sound bases for predicting short-term passenger demand. However, it also has several disadvantages, such as the relatively long collection time, the insufficiency to reflect the relationship between passenger behavior and ridership. Trip planner data, which emerged as a type of real-time transit information, could reduce the perceived waiting time of passengers and increase the transit ridership due to the improved satisfaction. Combining these two types of data could potentially cater to the interest of operators in matching the vehicle supply and passenger flow demand at an operational level. Our results show that it is novel and useful to incorporate trip planner data in short-term ridership prediction, however, entirely based on this kind of data would be inaccurate. Random Forest Regression outperforms the other six models that we have selected. The request-related features (variables) can take up 20% of the importance of short-term ridership prediction.