<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:dc="http://purl.org/dc/elements/1.1/" version="2.0">
  <channel>
    <title>OAR@UM Community:</title>
    <link>https://www.um.edu.mt/library/oar/handle/123456789/6641</link>
    <description />
    <pubDate>Sat, 21 Nov 2026 03:07:30 GMT</pubDate>
    <dc:date>2026-11-21T03:07:30Z</dc:date>
    <item>
      <title>A Bayesian hierarchical approach to modelling non-residential energy consumption in the Maltese islands</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/151209</link>
      <description>Title: A Bayesian hierarchical approach to modelling non-residential energy consumption in the Maltese islands
Abstract: The principal goal of this dissertation is to determine and analyse the factors&#xD;
influencing electricity consumption across non-residential buildings in the Maltese&#xD;
islands, while explicitly modelling the hierarchical dependencies present within the&#xD;
data. To achieve this, a series of statistical models are developed within a Bayesian&#xD;
framework and applied to non-residential electricity consumption data. The mod&#xD;
els are constructed in an order of increasing complexity. Specifically, the framework&#xD;
progresses from the first model which includes NACE and locality level random inter&#xD;
cepts to one that additionally includes covariates. This is followed by a model with&#xD;
added company-specific random intercepts and finally a formulation incorporating&#xD;
the pairwise interaction between NACE and locality. This incremental structure al&#xD;
lows for an evaluation of how additional complexity improves model performance and&#xD;
interpretability. Inference is conducted using Bayesian methods, providing full pos&#xD;
terior distributions for all parameters. Key summaries, including posterior means,&#xD;
medians and Highest Density Intervals, are used to quantify effects, associated uncer&#xD;
tainty and practical significance. Model adequacy is assessed through convergence&#xD;
diagnostics and graphical Posterior Predictive Checks to ensure that the fitted mod&#xD;
els provide a reliable representation of the observed data. Through comparative&#xD;
analysis of the specified models, the one offering the best data description and most&#xD;
reliable inference is identified. The most complex model is identified to be the opti&#xD;
mal specification, indicating that electricity consumption is not only influenced by&#xD;
sectoral and spatial factors, but also by their interaction. Substantial heterogeneity&#xD;
across sectors and regions, along with significant company-level effects, highlights&#xD;
the role of unobserved company-specific characteristics. Overall, the selected model&#xD;
quantifies the primary drivers of electricity consumption and explains the variance&#xD;
observed across different organisational levels. This dissertation highlights the value&#xD;
of Bayesian hierarchical modelling in addressing complex structured data and deep&#xD;
ens understanding of electricity consumption patterns among companies in Malta.
Description: B.Sc. (Hons)(Melit.)</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/151209</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>A hybrid ALNS-Tabu search algorithm for the airport stand and gate assignment problem</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/151207</link>
      <description>Title: A hybrid ALNS-Tabu search algorithm for the airport stand and gate assignment problem
Abstract: This dissertation addresses the Airport Stand and Gate Assignment Problem&#xD;
(ASGAP), a combinatorial optimisation problem that deals with the joint assign&#xD;
ment of parking stands and terminal departure gates to aircraft. While existing&#xD;
literature has treated stands and gates as the same physical resource, no prior work&#xD;
addresses the configuration in which all stands are remote, thereby requiring ex&#xD;
plicit coordination between stand and gate assignments under distinct feasibility&#xD;
constraints. Malta International Airport (MIA), which operates exclusively with&#xD;
remote stands, provides the motivating case study for this research.&#xD;
A Binary Non-Linear Program formulation is developed to minimise a weighted&#xD;
objective of passenger walking distance, aircraft taxiing distance, and a robust&#xD;
ness penalty penalising tight stand idle-time gaps. The ASGAP is proven NP-hard&#xD;
via a polynomial-time reduction from the Circular-Arc Graph Colouring Problem.&#xD;
Given this complexity, an Adaptive Large Neighbourhood Search with Tabu Search&#xD;
(ALNS-TS) metaheuristic is proposed, combining adaptive operator selection, an&#xD;
Exponential Monte Carlo acceptance criterion, and a tabu memory structure to&#xD;
diversify the search and avoid cycling.&#xD;
Computational experiments are conducted on four instances derived from MIA&#xD;
flight schedules, ranging from 45 to 115 aircraft. The Branch and Cut method ob&#xD;
tains optimal solutions for all instances and serves as a benchmark. The proposed&#xD;
ALNS-TS approach produces high-quality solutions within substantially shorter run&#xD;
times, achieving optimality gaps ranging from 9% to 17%. For the largest instance,&#xD;
ALNS-TS delivers assignments in under 15 minutes against the exact method’s 4.58&#xD;
hours to certify optimality. Overall, the proposed framework provides a practical&#xD;
and scalable framework for improving stand and gate coordination in airports oper&#xD;
ating under remote-stand configurations.
Description: B.Sc. (Hons)(Melit.)</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/151207</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>An adaptive large neighbourhood search framework for vehicle routing under stochastic travel times</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/151206</link>
      <description>Title: An adaptive large neighbourhood search framework for vehicle routing under stochastic travel times
Abstract: The Vehicle Routing Problem (VRP) is a fundamental optimization problem in&#xD;
logistics, where efficient routing decisions are required to reduce operational costs. In&#xD;
practice, travel times in VRPs are often uncertain due to congestion, accidents, and&#xD;
varying traffic conditions, which limits the realism of deterministic routing models.&#xD;
Therefore, this dissertation studies the Capacitated VRP with Stochastic Travel&#xD;
Times (CVRPSTT) and proposes a solution framework that combines data-driven&#xD;
scenario generation with metaheuristic optimization.&#xD;
Historical travel time matrices across 150 locations in Malta were first trans&#xD;
formed using Principal Component Analysis to reduce dimensionality, while pre&#xD;
serving the dominant variation of the data. The reduced data was then clustered&#xD;
using the k-means clustering technique, to generate representative traffic scenarios&#xD;
with associated probabilities. The CVRPSTT was then represented as a mixed in&#xD;
teger model with an expected objective function and solved using a hybridization of&#xD;
an Adaptive Large Neighbourhood Search (ALNS) and Simulated Annealing.&#xD;
The computational experiments on three instances of varying sizes show that&#xD;
the proposed framework consistently produces high-quality solutions within prac&#xD;
tical computational time. Among the different initialization methods considered,&#xD;
the Savings heuristic yields the strongest overall performance. Sensitivity analysis&#xD;
demonstrates that most of the selected parameters are robust, while moderate de&#xD;
struction levels and stronger search intensification generally improve solution qual&#xD;
ity. Comparisons with the Branch and Cut exact method on the smaller instances&#xD;
show that ALNS produces highly competitive solutions, while larger instances high&#xD;
light the practical limitations of exact approaches. Overall, the results demonstrate&#xD;
that integrating statistical learning techniques with adaptive metaheuristic search&#xD;
provides an effective and scalable approach for solving the CVRPSTT.
Description: B.Sc. (Hons)(Melit.)</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/151206</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>A game-theoretic bi-level optimisation approach to the competitive facility location problem</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/151204</link>
      <description>Title: A game-theoretic bi-level optimisation approach to the competitive facility location problem
Abstract: For companies that service customers through physical facilities, the location of&#xD;
each facility is often a critical factor in determining the success of that company. The&#xD;
Competitive Facility Location Problem is concerned with optimising the locations&#xD;
of the facilities belonging to multiple competing companies that operate in the same&#xD;
market. This dissertation takes several aspects of existing models and combines them&#xD;
in a unique way to form a novel variant, which is modelled as a Bi-Level Optimisation&#xD;
program using a Game-Theoretic lens. In particular, this dissertation focuses on a&#xD;
Leader-Follower budgeted variant where the value of the Follower’s budget is not&#xD;
known with certainty by the Leader. Further aspects included in the model are&#xD;
quality levels, elastic demands, pre-existing facilities, a maximum threshold on the&#xD;
number of facilities that can be owned, and customer behaviour based on the Huff&#xD;
gravitational rule.&#xD;
This dissertation investigates the computational complexity of solving Bi-Level&#xD;
Optimisation programs, and shows that exact and approximation methods may&#xD;
become impractical for large-scale problem instances. To address this, a memetic&#xD;
algorithm composed of Genetic Algorithm and Tabu Search is formulated, and a&#xD;
novel heuristic specifically adapted to our problem is proposed. These two algorithms,&#xD;
together with an exhaustive search exact method, are used to solve problem instances&#xD;
constructed to model the locations and types of electric vehicle charging stations in&#xD;
Sliema, Malta. The results show that both inexact methods significantly outperform&#xD;
the exact method in terms of solving speed. Moreover, the memetic algorithm is&#xD;
shown to consistently produce near-optimal solutions, whereas the heuristic method&#xD;
is faster but yields solutions of comparatively lower quality.
Description: B.Sc. (Hons)(Melit.)</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/151204</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
  </channel>
</rss>

