Aircraft operations require careful balance between maintaining sufficient crew availability and managing costs effectively. At KLM, current static methodologies for reserve crew planning achieve a 42.8% prediction accuracy on average, leading to either costly overstaffing or potentially disruptive understaffing. This research addresses this operational challenge by investigating the central question: ”How can we develop a self-learning model that enhances the prediction of reserve crew demand in the aviation industry by considering key factors such as seasonal variations and sickness absence patterns?”

Our literature review revealed a significant gap in current research. While studies have explored various aspects of airline operations and crew management, few have addressed the specific challenge of predicting reserve crew demand through machine learning approaches. The literature particularly lacked understanding of how multiple predictive factors interact and how illness patterns impact crew availability.

The research methodology involved systematic analysis of operational data from KLM spanning January 2013 through December 2023, focusing on the Boeing 777/787 fleet and Captain rank. We developed an integrated prediction framework comprising three main components: illness pattern analysis, reserve demand modelling, and operational validation. After evaluating multiple approaches including neural networks (LSTM, CNN) and traditional time series models, we identified an ARIMA(2,1,1) model with exogenous variables as optimal, based on both statistical criteria and operational requirements.

Analysis revealed previously undocumented relationships between operational factors, notably stronger correlation between incidental illness and reserve demand (r = 0.780) compared to structural illness patterns (r = 0.666). The developed model achieved 93% prediction accuracy, translating to conservative cost savings of €3.15 million in the first three quarters of 2024 through improved resource allocation and prevented flight cancellations. The model demonstrated particular strength in capturing dynamic patterns, successfully predicting major disruption events including an unexpected global illness wave in April 2024.
While the implementation shows promise for transforming reserve crew planning, several important challenges and areas for future research remain. The model’s reliance on historical patterns creates potential vulnerability to structural changes, and maintaining consistent data quality across systems presents ongoing challenges. Additionally, the impact of pilot preferences on reserve usage represents a complex system that warrants dedicated future research, as these preferences can significantly influence reserve crew utilization patterns. Despite these limitations, the system has reached implementation maturity, with KLM planning to utilize it for 2025 reserve capacity determinations.

This research establishes new possibilities for airline resource management while providing a framework for similar applications across the aviation industry. The findings suggest that machine learning approaches can significantly enhance prediction accuracy, though careful consideration must be given to operational constraints and data quality requirements. Future research opportunities include investigating more sophisticated approaches to handling extreme events, exploring the integration of additional predictive factors such as weather patterns and Not Fit to Fly data, and developing comprehensive models to account for pilot preferences and their impact on reserve crew demand.

In a growing and highly competitive online grocery market, online grocer Picnic must prioritize optimizing the efficiency of its logistics chain to achieve profitability. Last-mile delivery is a major contributor to operational costs, making its optimization essential. This thesis presents a study on optimizing last-mile delivery operations for Picnic through the application of a Hybrid Genetic Search (HGS) algorithm. The research addresses the Vehicle Routing Problem with Time Windows (VRPTW), factoring in Picnic’s unique operational constraints, including multi-compartment vehicles, a heterogeneous fleet, vehicle time windows, prioritized vehicles, and varied service times.

The study evaluates the performance of the HGS algorithm against Picnic's existing VROOM algorithm in real Picnic instances. Results demonstrate that, on average, HGS consistently outperforms VROOM, achieving significant reductions in total route duration. However, while HGS excels in optimizing route duration, it does not always minimize the number of vehicles used.

This research extends the state-of-the-art Hybrid Genetic Search (HGS) algorithm to tackle the specific challenges of Picnic’s vehicle routing problem. By applying this enhanced algorithm to a real-world setting, it bridges the gap between theoretical optimization advancements and practical industry implementation.

This thesis presents a sophisticated exploration into the use of genetic algorithms (GAs) for the design and optimization of golf courses, integrating both theoretical and practical perspectives. It outlines the challenges of traditional golf course design, including the strategic placement of features such as tees, greens, and hazards, and the need for efficient use of the landscape. By leveraging GAs, the research demonstrates a methodology to optimize these elements by simulating numerous design iterations, which helps in achieving an ideal balance between aesthetic appeal and playability. The core contribution of this
thesis lies in its detailed application of GAs to model and solve the complex problem of golf course layout with considerations for various design constraints like safety, play strategy, and land use efficiency. The results show that GAs can effectively generate diverse, innovative golf course designs that meet specified criteria and constraints, outperforming traditional methods in terms of the quality of the solutions, although at the cost of speed. This study not only enhances the toolkit available to golf course architects but also serves as a prototype for future research in the field. It suggests avenues for further exploration, including the integration of other computational techniques like reinforcement learning and advanced simulation tools, to further refine the design process and tailor golf course layouts to specific environmental and geographical conditions.

Cluster analysis in high dimensional data is a difficult but desirable task. Many existing methods fail to cluster high dimensional data due to what is known as the curse of dimensionality. Therefore, sophisticated clustering methods are in wide development. Along these lines, spectral modularity maximization emerged from the theory of random matrices and graph modularity. The method is based on a filtering of the spectral decomposition of similarity matrices. Despite the recent success of this method, we uncover a fundamental challenge of spectral modularity: the spectral modularity breaks down as the number of groups in a data set grows. To mitigate this challenge, we propose two solutions: one solution based on a regularization and one solution based on a normalization. We perform a thorough empirical analysis of the clustering performance of the solutions and find that, not only do our methods resolve the breakdown of spectral modularity, but they also outperform existing clustering methods in
a variety of settings.

This study introduces KarGus, a novel system for multi-document question answering (MD-QA) designed for diverse domains. KarGus integrates advanced Natural Language Processing techniques with Knowledge Graph (KG) construction and Graph Neural Networks (GNNs) to enhance retrieval performance across various specialized fields.
We explore the efficacy of combining semantic similarity, TF-IDF, and Named Entity Recognition features in KG construction and information retrieval. Experimental evaluation on a corpus of 30 documents (1810 pages, 10,853 text chunks) from corporate intelligence demonstrates that KarGus outperforms traditional embedding-based methods, achieving a Recall@5 of 0.850 compared to the baseline's 0.823 (p < 0.05). The optimal configuration emphasized semantic similarity (weight 0.75), keyword relevance (0.2), and entity information (0.05).
Analysis of the KG structure revealed moderately well-defined community structures and efficient information traversal properties. While GNN models showed promising training results, they underperformed in the retrieval task, highlighting challenges in GNN application to MD-QA.
This research contributes to the field of information retrieval by demonstrating the efficacy of integrating NLP techniques with graph-based approaches in MD-QA. The adaptable nature of KarGus suggests potential applications across various specialized domains. Future work will focus on validating cross-domain performance and refining GNN implementations for diverse retrieval tasks.

The combination of increasing global food demand with increased food security risks associated with climate change amid a decreasing number of skilled growers necessitates innovative solutions in green- house horticulture. Autonomous growing offers a solution based on greenhouse climate forecasting and (optimal) control. However, current theoretical models developed for greenhouse climate forecast- ing face limitations due to the in-depth physics knowledge required for their use and their dependence on intricate system parameters that are difficult to estimate. Conversely, machine learning models struggle with generalisation to unseen conditions and adherence to physical laws, leading to unrealis- tic predictions in greenhouse environments. This study addresses these challenges by exploring the use of Physics-Informed Machine Learning (PIML) techniques to enhance greenhouse climate fore- casting. A simple differentiable theoretical model for simulating the greenhouse climate is proposed to serve as prior physical knowledge of the system. Subsequently, a novel PIML model is introduced in the form of Controlled Aphynity (CA), which integrates insights from neural controlled differential equations and Aphynity, and is the first such model that allows for the augmentation of incomplete prior knowledge of a dynamical system with data-driven machine learning models while being adaptable to changing dynamics due to forces acting on the system. Experimental results show that CA is capable of successfully augmenting incorrect physics under changing dynamics on the task of humidity deficit prediction in the greenhouse. Furthermore, three ensemble methods combining CA with traditional ma- chine learning techniques are explored and demonstrate promising synergies. A detailed case study evaluates CA and the best-performing ensemble approach on the task of humidity deficit prediction under realistic greenhouse scenarios over a complete crop cycle. Both CA and the ensemble methods exhibit superior adherence to physical laws, lower data requirements, and improved performance on outliers compared to conventional machine learning methods. This research contributes to advancing greenhouse climate modelling, underscoring PIML’s potential in optimising the greenhouse climate.F

This thesis investigates the drivers of the cost per package of a major Dutch parcel corporation and develops a forecasting model to predict these costs accurately. Despite its efficiency in delivering millions of packages daily, a comprehensive understanding of the cost per package is lacking due to inconsistent definitions, unknown key drivers, and incomplete data. The primary objectives are to identify key cost drivers and develop a reliable forecasting model. Through the use of various statistical techniques and feature importance, volume and forecast realization ratio were identified as the most significant cost drivers. A comparative study of multiple forecasting models determined that the Least Squares model provides the best balance of accuracy and ease of implementation. The analysis highlighted the impact of accurate volume forecasts and the need for more detailed data collection and documentation. The research findings offer actionable insights for the company to optimize its operations and reduce costs. Additionally, this thesis contributes to the broader field of time series analysis and forecasting in logistics, serving as a case study for future research. Recommendations include improving volume forecasting accuracy, optimizing data collection processes, and establishing model monitoring. While the study's limitations include the dataset's size and granularity, future research could focus on creating more detailed datasets and exploring the impact of specific events on cost fluctuations and forecasting accuracy.

The branch-and-bound algorithm is used by solvers to efficiently find the optimal solution of discrete optimisation problems. It does so by sequentially partitioning parts of the search space based on the solution to the linear relaxation of the problem. This sequential decision-making is performed by the variable selection and node selection heuristics. The sequential nature of these heuristics makes them suitable for trajectory-based learning in the form of imitation learning and reinforcement learning. Learning problem-specific heuristics in this way has become increasingly popular in recent years. Despite their similarities, the two heuristics have very different dynamics during learning, and success has mainly been achieved for variable selection. In this work, we evaluate the node selection problem and formulate a learning to select paradigm for both imitation and reinforcement learning. We find that learning to select is generally more difficult due to the small margin of possible improvement over the current baselines, and the lack of informative features to distinguish nodes during ranking. These challenges are exacerbated by focusing on sibling comparisons, which are generally the most difficult due to the high similarity between the nodes. Sibling comparisons are also arguably the most important in node selection, however, due to the importance of plunging to reduce context switching overhead. The results indicate that both approaches fail to learn meaningful decision-making policies based on the limited fixed-size feature representation of the nodes. A code repository for reproducing and extending the experiments is publicly available at https://github.com/jgroenheide/rl2select.

UTURN aims to maximize the matching rate on its freight transport platform by efficiently connecting shippers with suitable carriers. To support this matching process, UTURN required a solution that was additive rather than restrictive on the platform. To achieve this, our research leverages recommendation systems and predictive models to help shipments find appropriate carriers in time while avoiding sending unnecessary email recommendations. We addressed three main questions: how to maximize matching probability through recommendations, how to balance the recommendation frequency to prevent spam, and how to ensure the solution adapts to market changes. We developed a recommendation system that ranks carriers based on their historical platform data using personalized k-nearest neighbour models and a custom similarity function. Tested in a controlled experiment, this system resulted in a 2.8% increase in the average matching rate, with improvements up to 3.4% in established regions and a peak of 6.2% in the Netherlands. However, the one-size-fits-all approach proved insufficient due to varying carrier capacities. To address the issue of deciding if and when to recommend, we introduced the Shipment Pulse Monitor (SPM), a decision tree-based predictive model that triggers recommendations when a shipment will likely be cancelled. To represent the trade-off between sending out recommendations to maximize matching probability and avoiding spam, we defined custom metrics for our problem. Our findings suggest that future work should focus on tailoring recommendations to specific carrier groups and combining hard cut-off points in data to allow a narrower focus for predictive models. Our work provides a comprehensive framework for improving matching rates on the UTURN platform with adaptive, targeted solutions.

The Dutch railway system is one of the most densely used systems worldwide and the busiest in Europe. Given the tight schedules, incidents can quickly cascade through the entire country if not handled properly. Alternative timetables are created to help train traffic controllers swiftly resolve such incidents. These schedules are currently created manually, but the team cannot keep up with demand. This is why CGI is developing VGB Solver, an application to automatically generate such timetables. The solver uses a branch and bound algorithm in which nodes are processed in best-first order, based on a heuristic value. For this thesis, different performance optimisations for this algorithm were implemented and analysed.

Various alternative formulations of the heuristic value, used to determine the likeliness of a node leading to a good solution, showed promising results.
One of these formulas resulted in solutions for 96% more scenarios than before, and improved the quality of solutions for other scenarios.

Using machine learning, a decision tree was created to predict whether applying another new formula for the heuristic value gives better results than the old formula. This classifier achieved an accuracy of 73.5% on the test data. Having the solver choose between the old and new formula based on that classification resulted in some scenarios with worse scores, but twice as many improved, and the average improvement was higher than the average deterioration.

Recommendations are made to conduct further experiments related to the heuristic value calculation. Furthermore, it is suggested to separate the scores for evaluating solution quality from the heuristic value formula, to facilitate more fine-grained changes to the calculation of the heuristic value.

Distribution Network Operators (DNOs) are confronted with a significant challenge to expand the capacity of the electricity distribution grid to facilitate the energy transition. Grid expansion planning for the distribution grid is a complex problem with many constraints and objectives. Earlier research has focused on metaheuristics such as genetic algorithms to optimise grid expansion designs in terms of resolved congestions, redundancy and costs. However, this algorithm is not used in practice by its intended users; grid architects. This research first identifies why the algorithm is not yet used and proposes several ideas to improve user adoption. Then a technical implementation of sketch-based optimisation using a novel shape similarity measure is presented. We investigate its behaviour with a realistic case study. Furthermore, a qualitative user study is done to evaluate the potential impact of sketch-based optimisation on distribution grid expansion planning. We conclude that the novel shape similarity measure enables sketch-based optimisation of distribution grid expansion planning and that it has potential to accelerate electricity grid design processes. Alliander, the largest DNO of the Netherlands, will implement sketch-based optimisation in their distribution grid design application later this year, making it available to 20+ grid architects to accelerate their design process.

n the rapidly evolving semiconductor industry, precise material characterization is crucial. This thesis focuses on enhancing Ion Beam Analysis (IBA), a pivotal tool in semiconductor characterization, through the integration of differential evolution optimization. This research proposes a single and multi spectra optimization algorithm approach controlled by a web application. Central to this work is a com- parative analysis of the proposed algorithms to simulated annealing and the DE algorithm proposed by Heller et al. [5]. This analysis shows a good performance of the proposed algorithms and a potential for industry application. The web application offers users a robust, user-friendly, and scalable interface for IBA optimization. By proposing and testing new IBA optimization methods, this thesis contributes significantly to semiconductor technology, offering new methods for material characterization at microscopic level.
This project is a joint project between Imec (Leuven, Belgium) and HZDR (Dresden, Germany).

This thesis proposes and develops an interface and model in which advanced optimisation for general employee scheduling is made available to non-experts in computer science or optimisation. The interface teaches, guides, configures, dynamically creates a constraint programming (CP) model, iteratively improves, decreases black box properties, increases trust in the outcome, and complies with relevant European Union Artificial Intelligence law. The objective of this study is to allow a wider range of organisations to take advantage of CP techniques, with the potential to greatly improve efficiency, reduce unfairness, meet company goals, and improve employee satisfaction.

Employees are assigned to personalised shifts based on the expected demands of departments within an organisation, these are set by the domain expert in the field of employee scheduling. Next, to tailor the model to the organisation’s needs, the domain expert is guided in setting both the restrictive assumptions and priorities of shift assignment. To optimise the generated optimal schedule, it is encouraged to create and compare multiple sets of configurations and subsequent schedules. Historical demand data, organisation structure, contract information, and employee preferences are included. Multiple visual design iterations have been made, after which a working interface has been developed and improved iteratively in conjunction with experts in the field. Validations with external domain experts from various industries and organisations have shown that the interface performed effectively in supporting the objectives. Further research can be done to improve the speed of solving, implement diversity of solutions, support for highly custom constraints based on natural language, or interface reusability for other optimisation problems.

Conventional fixed-line bus services are ineffective and costly in areas of low demand. Bus operators are motivated to scale down their service to reduce costs. Consequently, many rural areas in the Netherlands suffer from diminishing accessibility to jobs, hospitals, and education using public transport. Demand responsive transport (DRT) systems can offer a solution, but have been struggling in adoption by travellers because the need to make a reservation is perceived as an obstacle. Furthermore, the performance of DRT systems is complex and requires evaluation before implementation. In this context, this thesis considers the replacement of three low-demand bus lines by a DRT system in Voorne-Putten Rozenburg, Netherlands. Two scenarios are formulated: the replacement of a single bus line and the replacement of all three low-demand bus lines. The objective of this thesis is to evaluate the performance of a DRT system in serving the travel demand currently served by the bus lines. To obtain the travel demand, individual trips from the existing bus line were generated from aggregated data on boarding and alighting in 2023. The travel demand in a DRT system was evaluated using an agent-based simulation approach. To address the perceived obstacle of making reservations, the DRT system also considered on-demand requests and a proposed stop-based request where the user can request the DRT service physically at the bus stop with the consequence that the DRT system initially only knows the users' origin and not the destination. The main results suggest that a DRT system is an improvement over fixed-line buses in terms of waiting time and travel time, and CO₂ emissions when considering current travel demand in both scenarios. However, the costs are comparable. In the case that travel demand increases more than twofold, fixed-line buses are to be preferred. Considering the different request methods, only a small difference in user times was observed. This thesis concludes that replacing the bus lines with a DRT system can lead to an increase in accessibility. The proposed stop-based request method was shown to be a feasible alternative to the request methods that require the use of an app or phone call and could potentially lower the obstacle to using DRT.

The Multi-Objective Dial-a-Ride Problem (DARP) poses significant challenges in the field of transportation optimization, requiring the simultaneous optimization of conflicting objectives such as travel costs, emission, and customer ride times. In this research, we analyse two distinct approaches for tackling the Multi-Objective DARP: Mixed-Integer Linear Programming (MIP) solvers and Genetic Algorithms (GAs). Through a series of experiments and performance evaluations on diverse problem instances, we assess the strengths and weaknesses of each method. We compare their efficiency, scalability, and ability to generate Pareto-optimal solutions. Additionally, the study explores the impact of algorithmic variations on the convergence and solution quality of the Genetic Algorithm. The results demonstrate that MIP solvers seem entirely unsuited for the generation of quality Multi-Objective Pareto fronts. Of the Genetic Algorithms, the algorithm extended with our proposed guiding heuristics in its genetic operators, manages to construct the best quality Pareto front, outperforming the other algorithms in both finding the best objective solution values, as well as pareto diversity and convergence. We discuss the practical implications of our findings, offering recommendations for researchers and practitioners in the realm of transportation optimization and emission reduction.

The available technology to solve Mixed Integer Linear Programs (MILPs) has experienced dramatic improvements in the past two decades. Pushing this algorithmic progress further is essential for solving even more complex optimization problems that arise in practice. This thesis examines various methods to enhance Branch-and-Bound (B&B) based MILP solvers, focusing on areas such as branching and Machine Learning (ML) assisted rules. Through our analysis of current methodologies and the introduction of novel techniques, this thesis contributes to the development of more efficient and adaptive MILP solvers...@en

Lely is developing a tethered manure applicator robot, designed to work on the fields of farms. This vehicle called the Jojo, is constrained in its movement due to the attached tether. It can not make tight turns and can only reverse by backtracking the driven path. This thesis explores the path planning issue and provides a novel solution. The problem is best described as the shortest path from a starting point to a set of strokes that cover the field, with a constrained turning radius. We first explore existing solutions and discover that none exist that are suitable. Further research finds that the behaviour of this vehicle is best compared with steerable needles, but at a different scale. First, we construct a simulation environment for the vehicle. This simulation is used to develop the plan execution engine, path follower and safety checks. This is necessary for testing the solutions in the simulation and in the real world. Secondly, with some inspiration from the solutions provided for the steerable needles, we arrive at a method for finding a Jojo path. The method makes use of a large graph that represents translations associated with a given location. We also provide methods for attaching the starting point and destinations in the form of strokes, either via the start or end of the strokes or via the middle. A Jojo plan is then obtained by performing a directed Steiner tree approximation on this graph, where the starting point is the root and the terminals are the strokes. In order to make the approach fast enough, we spend a substantial amount of time profiling and analysing the program and design a number of performance improvements. These allow for the approach to solve instances that are larger than necessary. Thirdly, we test, compare and improve the approach. We look at both small instances to perform numerical analysis in order to optimise a set of parameters. We then inspect larger instances that are based on real world scenarios. Lastly, we perform a set of real world tests using the vehicle. One of these tests highlights a shortcoming of the algorithm, which we discuss and provide a solution for

This thesis aims at maximizing the profit of a strawberry producer while satisfying the retailer's demand and meeting other constraints. The amount of strawberries to be delivered to the retailer signed in the contract is the main decision variable to be optimized in the problem. Furthermore, the transportation scheduling is also optimized to help the producer reduce cost.

Inductive logic programming is a technique that generates logic programs which keep to a given specification using a background knowledge. We propose a new task in the field of pro- gram synthesis called Time-gated Partition-selection Inductive Logic Programming, consisting of splitting the background knowledge into partitions and selecting only the relevant partitions to a given set of examples. In order to show an initial direction of research and demonstrate the effectiveness of the approach, we have constructed a set of partitioning functions and a selection function. These were implemented using existing graph clustering and community detection algorithms applied to static call graphs of existing programs in the target language and using a linear time evaluation selection function. By comparing the inductive logic programming approach Popper to a version of Popper with its search space reduced using this technique, we show that these partitioning- and selection functions can improve the generated programs on three out of four different domains. Finally we show that there is a difference in partition quality by comparing the results to a random partitioning function. This work establishes background knowledge partitioning- and selection as a useful tool in program synthesis research.

Constraint programming solvers provide a generalizable approach to finding solutions for optimization problems. However, when comparing the performance of constraint programming solvers to the performance of a heuristic solver for an optimization problem such as cluster editing, the heuristic solver is able to find near-optimal and optimal solutions much faster. The goal of this research is to understand how the behaviour of such a heuristic solver can assist the performance of constraint programming solvers in optimization problems. In order to achieve this, first Chuffed, a state-of-the-art constraint programming solver was combined with a heuristic approach to cluster editing, with the goal of emulating the performance of the heuristic algorithm, in particular, being able to find near-optimal solutions faster. Continuing, the goal was to generalize the behaviour observed by the modified solver, by emulating the performance observed without the need for the specialized heuristic solver. The generalized approach is tested on a wide variety of different tests. An approach to value selection was developed that performs lookahead propagations for the two values of a boolean variable and selects the value that has the most optimal solution within the domain after performing the lookahead propagation. This approach added a significant time overhead that increased the overall solving time for many problems, with the lookahead configuration having a median increase of 8.7% over the default configuration for the optimization problems of the MiniZinc Challenge 2022. However, it was able to successfully emulate the performance of the heuristic solver, finding near-optimal solutions significantly faster than the default value selection. In particular, for the optimization problems of the MiniZinc Challenge 2022, on average, the lookahead configuration had a definite integral for the time vs objective graph 54.7% lower than the default Chuffed configuration.