Burak Boyaci | Lancaster University (original) (raw)

Papers by Burak Boyaci

arXiv (Cornell University), Jun 14, 2023

Motivation: Classification of gene trees is an important task both in analysis of multi-locus phy... more Motivation: Classification of gene trees is an important task both in analysis of multi-locus phylogenetic data, and assessment of the convergence of Markov Chain Monte Carlo (MCMC) analyses used in Bayesian phylogenetic tree reconstruction. The logistic regression model is one of the most popular classification models in statistical learning, thanks to its computational speed and interpretability. However, it is not appropriate to directly apply the logistic regression model to a set of phylogenetic trees with the same set of leaf labels, as the space of phylogenetic trees is not Euclidean. Results: It is well-known in tropical geometry and phylogenetics that the space of phylogenetic trees is a tropical linear space in terms of the max-plus algebra. Therefore, in this paper, we propose an analogue approach of the logistic regression model in the setting of tropical geometry. In our proposed method, we consider two cases: where the numbers of the species trees are fixed as one and two, and we estimate the species tree(s) from a sample of gene trees distributed over the space of ultrametrics, which is a tropical linear space. we show that both models are statistically consistent and bounds on the generalization error of both models are derived. Finally, we conduct computational experiments on simulated data generated by the multi-species coalescent model and apply our model to African coelacanth genomes to infer the species tree.

Transportation Research Board 90th Annual MeetingTransportation Research Board, Jan 23, 2011

Mobility and transportation are two of the leading indicators of economic growth of a society. As... more Mobility and transportation are two of the leading indicators of economic growth of a society. As cities around the world grow rapidly and more people and modes compete for limited urban space to travel, there is an increasing need to understand how this space is used for transportation and how it can be managed to improve accessibility for everyone. In a recent paper, Daganzo and Geroliminis (1) explored the connection between network structure and a network's MFD for urban neighborhoods with cars controlled by traffic signals and derived an analytical theory for the MFD using Variational Theory. Information needed to estimate this network MFD's are average network (total length of roads in lane-km, number of lanes, length of links), control (signal offsets, green phase and cycle time) and traffic (free flow speed, congested wave speed, jam density, capacity) characteristics. However in previous studies, Variational Theory has been applied only in cities with deterministic values of the above variables for the whole network and by ignoring the effect of turns. In our study we are aiming to generate an MFD for streets with variable link lengths and signal characteristics and understand the effect of variability for different cities and signal structures. Furthermore, this variability gives the opportunity to mimic the effect of turning movements and heterogeneity in drivers' behavior. This will be a key issue in planning the signal regimes such a way that maximizes the network capacity and/or the density range of the capacity.

Proceedings of the Genetic and Evolutionary Computation Conference Companion, Jul 7, 2021

North West Air Ambulance (NWAA) provides helicopter emergency medical service to the Northwest of... more North West Air Ambulance (NWAA) provides helicopter emergency medical service to the Northwest of England. Their three healthcare teams provide their service from two bases with three helicopters. They face some research questions to understand the impact of the air ambulance base locations and the healthcare teams' composition on their services. This paper aims to address those questions by modelling their operations into a location-related decision problem. Then we developed a matheuristic approach to solving the model and generated many realistic instances from historical data to validate our proposed approach's robustness. With the help of our experimental results, we examine the e ect of adjusting air ambulance base locations as well as team con gurations on the service quality measures to answer the research questions. CCS CONCEPTS • Computing methodologies → Randomized search.

Transportation Research Part B-methodological, Dec 1, 2012

Recent experimental analysis has shown that some types of urban networks exhibit a low scatter re... more Recent experimental analysis has shown that some types of urban networks exhibit a low scatter reproducible relationship between average network flow and density, known as the macroscopic fundamental diagram (MFD). It has also been shown that heterogeneity in the spatial distribution of density can significantly decrease the network flow for the same value of density. Analytical theories have been developed to explore the connection between network structure and an MFD for urban neighborhoods with cars controlled by traffic signals. However these theories have been applied only in cities with deterministic values of topological and control variables for the whole network and by ignoring the effect of turns. In our study we are aiming to generate an MFD for streets with variable link lengths and signal characteristics and understand the effect of variability for different cities and signal structures. Furthermore, this variability gives the opportunity to mimic the effect of turning movements. Route or network capacity can be significantly smaller than the capacity of a single link, because of the correlations developed through the different values of offsets. The above analysis would not be possible using standard traffic engineering techniques. This will be a key issue in planning the signal regimes such a way that maximizes the network capacity and/or the density range of the maximum capacity.

Procedia - Social and Behavioral Sciences, 2011

As more people through different modes compete for the limited urban space that is set aside to s... more As more people through different modes compete for the limited urban space that is set aside to serve transport, there is an increasing need to understand details of how this space is used and how it can be managed to improve accessibility for everyone. Ultimately, an important goal is to understand what sustainable level of mobility cities of different structures can achieve. Understanding these outcomes parametrically for all possible city structures and mixes of transport modes would inform the decision making process, thereby helping cities achieve their sustainability goals. In this paper we focus on the network capacity of multimodal systems with motorized traffic and extra emphasis in buses. More specifically, we propose to study how the throughput of passengers and vehicles depends on the geometrical and operational characteristics of the system, the level of congestion and the interactions between different modes. A methodology to estimate a macroscopic fundamental diagram and network capacity of cities with mixed-traffic bus-car lanes or with individual bus-only lanes is developed and examples for different city topologies are provided. The analysis is based on realistic macroscopic models of congestion dynamics and can be implemented with readily available data.

Transportation Research Part B-methodological, Dec 1, 2019

In this paper, we study the operations of a one-way station-based carsharing system implementing ... more In this paper, we study the operations of a one-way station-based carsharing system implementing a complete journey reservation policy. We consider the percentage of served demand as a primary performance measure and analyze the effect of several dynamic staff-based relocation policies. Specifically, we introduce a new proactive relocation policy based on Markov chain dynamics that utilizes reservation information to better predict the future states of the stations. This policy is compared to a state-of-the art staff-based relocation policy and a centralistic relocation model assuming full knowledge of the demand. Numerical results from a real-world implementation and a simulation analysis demonstrate the positive impact of dynamic relocations and highlight the improvement in performance obtained with the proposed proactive relocation policy.

With production having come to an end at the Sellafield nuclear site in West Cumbria, focus is no... more With production having come to an end at the Sellafield nuclear site in West Cumbria, focus is now turning to its decommissioning, and the safe clean-up of legacy nuclear waste. The decommissioning project is expect to take in excess of 100 years to complete and cost over $90 billion. Given the large scale and complexity of this project, it is crucial that each task is systematically choreographed according to a carefully designed master schedule, determining both the start time, and duration of each activity, in order to reduce the project completion time. We show how this scheduling problem can be formulated as a continuous time/resource trade-off problem with a single renewably constrained resource, and subsequently propose a heuristic approach to constructing such a master schedule.

Although they have different objectives, emergency response systems and on-demand transportation ... more Although they have different objectives, emergency response systems and on-demand transportation systems are two similar systems in the sense that both deal with stochastic demand and service time which create congestions for moderate level of demand. Emergency response system location problems are one of the early problems immensely dealt in the literature. These problems are modeled by either set covering or transportation models which do not give much attention to the stochastic nature of the problem. On-demand transportation is a newly developing type of transportation system and literature is not broad enough but has similarities with emergency response systems. In this research, our aim is to solve facility location problem with stochastic demand and service time. Specifically we are dealing with temporal and spatial stochasticity which emerge because of the uncertainty in demand and service time. Recently we have developed a mixed aggregate hypercube model which are extensions to Larson (1974) and Boyacı and Geroliminis (2012). Results are promising and applicable to real life instances.

In spatial queues, servers travel to the customers and provide service on the scene. This propert... more In spatial queues, servers travel to the customers and provide service on the scene. This property makes them applicable to emergency response (e.g. ambulances, police) and on-demand transportation systems (e.g. paratransit, taxis) location problems. However, in spatial queues, there exist a different service rate for each customer-server pairs which creates Markovian models with enormous number of states and makes these approaches difficult to apply on even medium sized problems. Because of demand uncertainty, the nearest servers to a customer might not be available to intervene and this can significantly increase the service times. In this paper, we propose two new aggregate models and an approximate solution method with a dynamic programming heuristic. Results are compared with existing location models on hypothetical and real cases.

We study the operations of station-based one-way carsharing systems that enforce a complete journ... more We study the operations of station-based one-way carsharing systems that enforce a complete journey reservation policy. Under such regulation, users are required to reserve both a vehicle at the origin station and a parking spot at the destination station whenever they wish to make a trip. Reservations can be made up to one hour in advance and users do not have to specify in advance the exact pickup and drop-off times. These attractive customer-oriented rental conditions guarantee the availability of vehicles and parking spots at the start and end of the customers' journeys but may result in an inefficient use of resources. Notwithstanding, reserved vehicles/parking spots provide information about resources that are about to become available. In this work, we develop a Markovian model for a single station that explicitly considers journey reservation information and estimates the expected near future demand loss using historical data. The output of the model is integrated in a new proactive dynamic staff-based relocation decision algorithm. The proposed algorithm was tested in the field on the Grenoble car-sharing system and compared to other dynamic and static approaches. Real-world results are reinforced by an extensive simulation experiment using real transaction data obtained from the same system.

In this research, we are modeling the facility location problem for systems with many servers and... more In this research, we are modeling the facility location problem for systems with many servers and stochastic demand, such as emergency response and on-demand transportation systems, by using spatial queues,specifically hypercube queuing models (HQMs). We are proposing two extended HQMs and a heuristic method which uses an approach similar to dynamic programming. Briefly, the heuristic solves a subproblem for each sub region which is generated by using similarities in demand and then combines them accordingly. Preliminary analysis gives promising results for some real life instances.

Transportation Research Board 91st Annual MeetingTransportation Research Board, Jan 22, 2012

Different than the conventional queueing systems, in spatial queues, servers travel to the custom... more Different than the conventional queueing systems, in spatial queues, servers travel to the customers and provide service on the scene. This property makes them applicable to emergency response systems (e.g. ambulances, police, fire brigades) and on-demand transportation systems (e.g. shuttle bus services, paratransit, taxis). The difference between the spatial queues and conventional queueing systems is various types of customers and servers and different service rates for different customer-server pairs. For the Markovian arrival and service characteristics, one of the methods to find system performance measures is to model and calculate steady state probability of the Markov chain for the hypercube queueing model (HQM) (Larson, 1974). One of the obstacles on the way to apply HQMs to real life problems is the size of the problem; it grows exponentially with the number of servers and a linear system with exponential number of variables should be solved for each instance. In this research, in order to increase scalability of the problem, we propose two new models. In addition to that, we modeled the problem by using Monte Carlo simulation and tested the convergence and stability properties of the simulation results and compare them with stationary distributions.

The rapid growth expected in the offshore wind sector means there is an increasing opportunity to... more The rapid growth expected in the offshore wind sector means there is an increasing opportunity to find savings from conducting operations and maintenance activities more efficiently. The predicted increase in the size and quantity of offshore wind farms will require industry to have access to mathematical tools for scheduling maintenance activities in order to exploit potential savings fully. In order to complete a maintenance activity, a pre-specified combination of skilled personnel, equipment and vessel support is required for a specific duration at the location of the task. A heterogeneous fleet of vessels is typically responsible for transporting physical assets around the wind farm and conducting personnel transfers. Vessel movements must also satisfy any limitations in wind turbine accessibility, in conjunction with safety constraints imposed by offshore weather conditions. In this research, we have created a mathematical model of the problem that is capable of determining the best routes for vessel movements and the ideal times to undertake crew transfers. Our approach can compute high quality schedules that minimise the costs of performing both corrective and preventive maintenance tasks, whilst accounting for lost production. We illustrate our results on a test wind farm. Our solutions demonstrate the possible benefits from splitting pick-up and drop-off operations across different vessels and only performing a subset of tasks within a task intensive environment. It is possible to extend our model to incorporate a set of scenarios that represent the stochastic evolution of weather and sea conditions in future shifts. Solving the resulting model with a rolling horizon approach allows us to produce a detailed solution for the current shift, which contains actions informed by the relative likelihoods of future weather patterns.

The capacitated multi-facility Weber problem is concerned with locating m facilities in the plane... more The capacitated multi-facility Weber problem is concerned with locating m facilities in the plane, and allocating their capacities to n customers at minimum total cost, which is a non-convex optimization problem and difficult to solve. In this work we relax the capacity constraints and solve the uncapacitated Lagrangean subproblems every step of a subgradient algorithm. Their objectives have facility dependent distance functions, which is different than the usual multifacility Weber problem. We solve them by branch-and-price using column generation with concave minimization pricing.

The rapid growth expected in the offshore wind sector presents a growing opportunity to find savi... more The rapid growth expected in the offshore wind sector presents a growing opportunity to find savings from conducting operations and maintenance activities more efficiently. The predicted increase in the size and quantity of offshore wind farms means that mathematical tools for scheduling maintenance activities will be necessary to exploit economies of scale fully. In order to complete an activity, a predetermined combination of skilled personnel, equipment and vessel support is required to be present at its location for the duration of the task. A fleet of heterogeneous fleet of vessels is responsible for both transporting the resources around the wind farm and conducting personnel transfers. Vessel movements must also account for limitations imposed by offshore weather conditions and the periodic need to return resources to port. In this research, we have developed a mathematical model capable of determining the best routes for vessel movements and the ideal times to undertake crew transfers. Our mixed-integer programming formulation can compute high quality schedules that minimize the twin costs of performing maintenance and lost production. We extend our optimization model to include a set of scenarios that represent the stochastic evolution of weather and sea conditions in future shifts. Solving the resulting model with a rolling horizon approach allows us to produce a detailed solution for the current shift, which contains actions informed by future weather patterns.

Networks, 2021

Matchings and T‐joins are fundamental and much‐studied concepts in graph theory and combinatorial... more Matchings and T‐joins are fundamental and much‐studied concepts in graph theory and combinatorial optimization. One important application of matchings and T‐joins is in the computation of strong lower bounds for arc routing problems (ARPs). An ARP is a special kind of vehicle routing problem, in which the demands are located along edges or arcs, rather than at nodes. We point out that the literature on applying matchings and T‐joins to ARPs does not fully exploit the structure of real‐life road networks. We propose some ways to exploit this structure. Computational results show significant running time improvements, without deteriorating the quality of the lower bounds.

Fierce competition between airlines has led to the need of minimising the operating costs while a... more Fierce competition between airlines has led to the need of minimising the operating costs while also ensuring quality of service. Given the large proportion of operating costs dedicated to aircraft maintenance, cooperation between airlines and their respective maintenance provider is paramount. In this research, we propose a framework to develop commercially viable and maintenance feasible flight and maintenance schedules. Such framework involves two multi-objective mixed integer linear programming (MMILP) formulations and an iterative algorithm. The first formulation, the airline fleet maintenance scheduling (AMS) with violations, minimises the number of maintenance regulation violations and the number of not airworthy aircraft; subject to limited workshop resources and current maintenance regulations on individual aircraft flying hours. The second formulation, the AMS with tail assignment (TA) allows aircraft to be assigned to different flights. In this case, subject to similar constraints as the first formulation, six lexicographically ordered objective functions are minimised. Namely, the number of violations, maximum resource level, number of tail reassignments, number of maintenance interventions, overall resource usage, and the amount of maintenance required by each aircraft at the end of the planning horizon. The iterative algorithm ensures fast computational times while providing good quality solutions. Additionally, by tracking aircraft and using precise flying hours between maintenance opportunities, we ensure that the aircraft are airworthy at all times. Computational tests on real flight schedules over a 30-day planning horizon show that even with multiple airlines and workshops (160 0 0 flights, 529 aircraft, 8 maintenance workshops), our solution approach can construct near-optimal maintenance schedules within minutes.