Congratulations to our winners this year. The SLA Student Awards 2024 reached new goals this year with nearly 30 entries from the following universities: University of Leeds, UCL, University of Greenwich, University of Nottingham, University of Sheffield. Thanks to everyone for sharing and encouraging students to enter.
We awarded two prizes in each category; Masters and Undergraduates.
Masters Award winners 2024
Our two Masters winners were:
- James Tan, UCL
- Feruza Kachkinbayeva, University of Greenwich
Undergraduate Award Winners 2024
Taking home the prizes in the Undergraduates category were:
- Freddie Wallace, University of Leeds
- Asli Doga Kanturk, University of Greenwich
Highly Commended
As mentioned above, the judges were impressed with the high standard of entries this year and also wanted to mention four other submissions which deserved a shout out. Well done to our High Commended students:
Undergraduate
Firstly two students, both from the University of Leeds
- Charlotte Taylor
- Awais Hafesji
Masters
And finally, from our Masters category, two students from University College London:
- Caroline Beck
- Nikhil Desai
Meet our winners
Masters Award 2024
1st Place – James Tan, University College London
Dissertation Title: ‘Sky’s The Limit (or 120m): Urban Drone Delivery with Agent-Based Modelling’
Dissertation Synopsis: Ride-hailing services and urban deliveries are expected to grow in cities, contributing to congestion and other negative externalities. Autonomous drone deliveries are viewed as a potential solution to alleviate congestion, more environmentally sustainable and cost effective. However, integration of such technologies in cities has been slow due to technological limitations, public acceptance, economic feasibility and regulatory challenges. Hence, this study focuses on how the transition from a single to multiple delivery drone companies impacts the potential for aerial collisions in urban environments, particularly as the maximum number of allowable
drones increases. The research aims to provide a framework for the proliferation and regulation of drone operations in cities, addressing a notable gap in existing literature which often prioritise cost or time optimisation for operators over safety concerns relevant to regulators. Utilising an Agent-Based Model created in Unity, this research simulates drone delivery scenarios using OpenStreetMap data for a subset in
London. Five submodels manage various aspects of the simulation, including generating delivery points, truck placement with non-overlapping and overlapping clusters to simulate single and multiple drone companies scenario respectively, drone movement and avoidance manoeuvres. Key metrics analysed include average 3 unique drone encounters, average avoidance manoeuvres and percentage time in horizontal or vertical proximity to other drones. Results demonstrate that the transition from single to multiple drone companies creates a more chaotic airspace
as potential collisions increases. The introduction of additional trucks generally assists in reducing potential collisions, though exceptions exist. The findings suggest an urgent need for Urban Traffic Management System to ensure a structured and safer airspace to complement onboard collision avoidance mechanisms. The study also proposes avenues for future research such as improved algorithms for the single and multiple companies scenarios and comparative analysis of hybrid dronetruck models versus standalone operations.
Research Finding
- The transition from single to multiple drone companies in a city will create a far more chaotic airspace. Even the ‘worst’ configuration of number of drones in the sky and number of trucks as mobile launch points for the single company scenario is still better than the ‘best’ configuration for the multiple companies scenario in terms of potential aerial collisions.
- Including more trucks as mobile launch points can help to reduce potential aerial collisions, although they need to be spaced out adequately as well.
- Policy implications include that regulators could explore 1) a universally accepted metric on drone density and acceptable threshold in cities for regulating safer skies, 2) policies to optimise the number and placement of depots/mobile trucks for drone delivery operations to reduce drone density in the area and 3) creation of predetermined air corridors for a more orderly airspace.
- Drone delivery is a complex field with many potential future research direction (e.g. consideration of no-fly zones, optimal drone pathfinding and collision avoidance algorithm, (mobile) launch site placements, consideration of economic, environmental, social and safety impacts, etc.)
Further comments
The study area of the agent-based model was a subset of Islington and Camden. It was created in Unity, an engine for game development but also used by the government and aerospace sector for modelling, visualisation, simulation and machine learning (code available at https://github.com/jamestansongen/casa_msc_dissertation). In the top image, the drone is moving towards target and in the bottom image changing direction using artificial potential fields (i.e. repulsion force) to avoid other drone in avoidance radius. During this process, they record that a potential collision could have occurred if no evasive
manoeuvre was taken
Watch James’s presentation
Contact James Tan
2nd place – Feruza Kachkinbayeva – University of Greenwich
Dissertation Title: ‘Location Intelligence for Café Site Selection: Predicting Success and Commercial Rent Prices in London‘
Dissertation Synopsis: In London’s fast-paced and competitive café market, selecting the right location is one of the most crucial decisions for success. This dissertation addresses the core challenge entrepreneurs face: how to identify locations that offer both high business potential and financial viability. By integrating advanced spatial analytics with decision-making models, this research delivers a powerful tool for café site selection that not only predicts success but also uncovers cost-effective opportunities.
Focusing on London’s Lower Super Output Areas (LSOAs), the study analyzes a comprehensive set of critical success factors, including demographic affluence, accessibility to public transport, local economic conditions, competition density, and proximity to key amenities. Clustering techniques and the Analytic Hierarchy Process (AHP) are employed to systematically prioritize these factors, pinpointing the most promising areas for café establishments.
Predictive models, constructed using the effective Random Forest algorithm, precisely estimate commercial rent prices for each LSOA. This data, sourced primarily from open public government websites, ensures the transparency and reliability of the information.
A standout feature of this research is its comparative analysis of predicted rents, actual market data, and café success metrics. This dual-framework approach identifies locations where rents are lower than expected but have high business potential, providing entrepreneurs with a strategic advantage in site selection. This method goes beyond mere predictions, highlighting where businesses can thrive while minimizing rent costs, thus making the decision-making process both data-driven and financially prudent.
Additionally, the study includes interactive geographic visualizations that make complex data accessible and actionable for stakeholders. These tools enable entrepreneurs, urban planners, and business developers to visualize optimal café locations, ensuring decisions are backed by solid, real-world insights. This study merges academic research with practical applications to transform café site selection in London, enhancing predictive accuracy and providing cost-effective solutions for entrepreneurs in a competitive market.
Main Findings
| Throughout the research process, numerous challenges were navigated, leading to deeper insights into the complexities of urban economic patterns. A pivotal finding was the correlation between various demographic and economic factors—such as income levels and proximity to public transport—and both café success and rent costs. This analysis showed that locations with better accessibility and higher foot traffic not only correlate with higher potential for café success but also significantly influence rent prices. These factors provide a unique market advantage for new café entrepreneurs. Interactive Geographic Visualizations: This research has yielded a series of interactive geographic visualizations that are vital for demonstrating the practical applications of the findings. These tools simplify the complex interplay of demographics, rent, and business success metrics, making them accessible and actionable. The visualizations serve as essential resources, aiding stakeholders in making informed, data-driven decisions that enhance both the profitability and viability of new café locations. Here are the visual tools that can be explored: |
| London Café Success Map: Highlights areas with high potential for café success based on comprehensive analysis. London Predicted Rent Map: Visualizes areas with expected rent prices, aiding in financial planning and budgeting for new cafés. London Café Success and Rent Map: Combines insights from both the success and rent maps to identify optimal locations for new establishments. |
By emphasizing these correlations and providing predictive insights, these maps facilitate strategic decisions, not just predicting where cafés might succeed but also identifying where they can achieve success with optimal financial efficiency. This integration of data transforms these insights into powerful tools for urban planning and business development
Final thoughts:
The methodology established in this study is not only applicable to café site selection but also reveals potential for broader urban analytics applications. The use of open-source government data combined with advanced modeling techniques presents a cost-effective approach for scaling urban analysis across different sectors. Personal insights gained from this research emphasize the importance of accessibility and local economic dynamics in urban development. These findings inspired a proposed framework for policymakers on integrating data-driven strategies into urban planning, which could help optimize resource allocation and enhance public infrastructure planning. Furthermore, this research journey has underscored the need for ongoing collaboration between academic researchers, local governments, and industry stakeholders to refine urban analytical tools and ensure they meet real-world needs effectively.
Engaging with this interdisciplinary research has been a transformative experience, highlighting the crucial role of innovation and adaptability in addressing urban challenges. It has also shaped my approach to data analysis, making me a better-equipped researcher and problem-solver.
Contact Feruza
Undergraduate Award 2024
1st place – Freddie Wallace, University of Leeds
Dissertation Title: ‘Identifying Key Drivers of Closures in a Multichannel Era of Banking in England and Wales.’
Dissertation synopsis –
Rationale: Individuals must have good access to financial services to participate in the economy and accumulate wealth (Dymski, 2005). An uneven distribution of access to financial services can therefore create wealth inequalities and must be investigated. In the case of the UK, access has become a particularly prevalent topic. This is due to an estimated 56.4% of bank branches closing between 2015 and the end of the first quarter of 2024 (Geolytix, 2024). However, this has been somewhat compensated for by the rise in internet banking usage, and therefore a multichannel analysis is required.
Objective: The dissertation investigates the key predictors of bank closure in England and Wales. The analysis incorporates spatial, socioeconomic and internet usage data, to provide a landscape of banking access in a multichannel era. Finally, it uses this context to outline policy implications.
Method: The research adopts a novel approach to stepwise regression. This involves creating a series of binomial logistic regression models with a variety of predictor variables. These models predict whether each bank operating in 2015 will close by the end of the first quarter of 2024. The models build in complexity by adding a category of variables each time whilst avoiding multicollinearity. Eventually a backwards stepwise approach is taken to simplify the model which highlights only the key variables influencing bank closure. Finally, the Internet User Classification (IUC) is added to this simplified model and its effects are reviewed.
Discussion: The research highlights the risk of financial exclusion, with individuals who are faced with the highest rates of bank closures being unable to effectively switch to online banking due to poor digital literacy. Actionable insights are provided for policymakers and location planners to ensure that individuals can access the value of financial services through either a physical or digital channel.
Key findings
The research was able to correctly predict 72% of bank closures using spatial, demographic and banking data from GEOLYTIX. It also derived the following from coefficient estimates:
- East Midlands, London and Yorkshire on the Humber were all more likely to experience bank closures than the South East (intercept).
- As distance to competition or a Post Office increases, the chance of closure also increases. Conversely, an increased distance to another home bank very minimally decreases the chances of closure.
- As the percentage of a bank’s catchment population in the social grade DE rises, the likelihood of bank closure falls significantly.
- Banks classed as being suburban are generally less likely to experience closure.
- The IUC was found to be a highly significant predictor of bank closure, and its inclusion was able to improve the fit and accuracy of the simplified model. This research found that those with the least internet engagement were the most likely to experience bank closure.
These findings recognise the considerations of location planners on the impacts of bank closure, since branches are closing at a slower rate where individuals are already of a low socio-economic status, and where they might have greater difficulty in finding alternative branches (suburbia). However, it also highlights the need for a broader consideration of datasets, since those most heavily impacted are the least likely to have the ability to switch to online banking, increasing the risk of financial exclusion.
The increased application of a dataset like the IUC will minimise risk, as the catchment population of closed branches are more likely to be willing and able to migrate to online banking to seek a relative level of access to financial services. Therefore, the dissertation recommends that the IUC or an equivalent dataset is implemented both in location planning teams and regulatory policy.
Further comments
The models were constructed as follows, with colours denoting variable categories (orange: spatial, green: demographics, blue: bank characteristics):
Figure 1: Model 1, with only spatial variables. Model 2, with spatial and demographic variables. Finally Model 3, with spatial, demographic and bank characteristic variables.
Figure 2: The backwards stepwise process simplified Model 3 into Model (No IUC). Finally, the IUC is added to create the Model (IUC).
Watch Freddie’s Presentation
Contact Freddie Wallace
2nd place – Asli Doga Kanturk, University of Greenwich
Dissertation Title: ‘Urban Computing and Planning: Utilising Geo-Tagged Social Media Data for Identifying Leisure Hubs‘
Dissertation Synopsis:
This dissertation investigates the innovative application of geo-tagged social media
data for urban planning, with the primary aim of identifying and analysing leisure
hubs in London. The research leverages urban computing techniques and advanced
social media analytics to uncover spatial patterns and public perceptions associated
with leisure activities across diverse urban neighbourhoods.
The research rationale is grounded in the need for real-time, data-driven insights into urban dynamics to support more effective and targeted urban planning. By harnessing data from the Foursquare API, the study provides a rich dataset of geotagged social media interactions. This data is meticulously validated and processed using QGIS to ensure its accuracy and integrity, forming the basis for further analysis.
Machine learning models, including SVM, Naive Bayes, and Random Forest, are employed to classify leisure activities into categories such as dining, shopping, and cultural activities. Random Forest is identified as the most effective model, achieving high accuracy and reliability. Sentiment analysis using TextBlob is conducted to gauge public emotions and perceptions of various leisure hubs.
The direction of the research is operationalised through an interactive map interface developed with Folium, which displays real-time data and allows users to filter by borough, neighbourhood, and leisure category. This interface integrates additional geographical layers, such as town centres and opportunity areas, to provide a comprehensive contextual understanding.
The application also generates detailed reports based on user-selected filters, offering visual summaries and actionable insights for urban planners, policymakers, business owners, and residents. These reports compile data on leisure activity patterns, public sentiment, and geographic context, facilitating strategic planning and decision-making
Ultimately, the project underscores the potential of integrating real-time, geo-tagged social media data into urban planning. It aims to enhance the accuracy and relevance of urban planning, promoting vibrant, inclusive, and engaging urban environments. The research highlights the importance of big data and advanced analytics in modern urban planning, supporting more informed and effective decision-making.
Key findings
Data Validation and Pre-Processing:
Using QGIS, I validated and cleansed GIS datasets to ensure integrity and accuracy, crucial for seamless integration with the geospatial framework.
Text Classification and Machine Learning:
Implemented three machine learning models (SVM, Naive Bayes, and Random Forest) for classifying leisure activities. Random Forest had the highest accuracy (85.71%), precision (0.86), recall (0.86), and F1 score (0.84), proving reliable for categorizing leisure activities.
Sentiment Analysis and Geospatial Insights:
Conducted sentiment analysis using TextBlob to gauge public perceptions of leisure hubs across London. Positive sentiments were high in areas with dining and parks in Greenwich. Southwark’s cultural offerings and nightlife also received positive feedback, reflecting the area’s vitality.
Geospatial Analysis Outcomes:
Utilized Folium for interactive mapping, highlighting clusters of leisure activities across London. The integration of additional geographical layers such as town centres, opportunity areas, and central activity zones provided a comprehensive view of leisure dynamics. QGIS technologies were employed to correct and validate these layers, ensuring they accurately represented the urban landscape and
provided meaningful insights
Practical Applications:
This application is highly practical and user-friendly for various stakeholders:
- Urban Planners: Can use it to understand leisure dynamics, address negative feedback, and make informed decisions about infrastructure investments.
- Business Owners: Those looking to open new shops can identify promising locations based on public sentiment and activity patterns.
- Families: Individuals seeking to move can explore neighborhoods and understand their dynamics, ensuring they choose areas that match their preferences and lifestyle needs.
- Policy Makers: Can utilize the application to develop targeted strategies that enhance urban life and address community concerns.
These findings demonstrate the potential of integrating social media analytics into urban planning, providing a data-informed approach that benefits various users and enhances urban environments.
Key Features
Below is an overview of the application’s key features:
- Interactive Map Interface: Users can filter by borough and neighbourhood
to view leisure hubs, categorised by activities like dining, shopping, and
nightlife. This allows planners to explore specific areas in detail. - Geographical Layer Integration: The map includes layers such as town
centres, opportunity areas, and central activity zones, providing context to
the leisure hubs and aiding comprehensive urban planning. - Detailed Venue Information: Clicking on a pin reveals detailed venue
information, including name, address, and user reviews. Sentiment analysis
of reviews provides insights into public perception, helping planners
improve areas with negative feedback. - Leisure Hub Distribution: Analytical tools visualize the distribution of leisure hubs by category, with bar charts showing the frequency of activities like dining and shopping. This helps identify trends and inform urban development strategies.
- Comprehensive Reporting: The application generates detailed reports
with visual summaries, tailored to user selections (e.g., borough,
neighbourhood, leisure category). These reports provide actionable data for strategic planning and policy-making.
These features collectively provide a powerful tool for enhancing the
liveability and vibrancy of urban environments, offering visual and datadriven insights crucial for informed, strategic urban development decisions
Contact Asli Doga Kanturk
Congratulations to all our winners, we had such a great response of very good quality dissertations this year.