Linus Dietz

Lecturer (Assistant Professor) in Computer Science Education

Research Projects

Urban Parks for Promoting Health

In collaboration with Sanja Šćepanović, Ke Zhou, Daniele Quercia (Nokia Bell Labs, UK), and André Felipe Zanella (Telefónica, Spain)

Urban parks are essential for promoting health through recreation and leisure, but their specific roles in supporting diverse health-enhancing activities have been underexplored. Traditional studies often focus on size, greenery, or accessibility, overlooking how individual park spaces and elements contribute to health. To address this, we developed a taxonomy of six health-promoting activity categories—physical, mind-body, nature appreciation, environmental, social, and cultural—and assessed parks in 35 global cities using data on 23,477 parks and 827,038 features from OpenStreetMap.

Our analysis revealed distinct trends: North American parks prioritize physical activities, while European parks favor nature appreciation. Parks in city centers typically offered more diverse activities compared to suburban ones, and significant disparities in park standards were evident across cities. While Tokyo and Paris showed uniform park quality, cities like Copenhagen and Rio de Janeiro had more pronounced inequalities. These findings underscore the need for equitable, health-promoting urban park designs.

In cities with extensive park systems, like London, we explored how recommender systems could help residents discover parks that match their leisure and health preferences. Using data from a resident survey and over a million geotagged Flickr images, we found that proximity-based recommendations often fell short, favoring popular parks over lesser-known ones. Personalized models effectively addressed these biases, broadening the range of parks highlighted and better aligning with user preferences. This research highlights the potential of recommendation systems to connect urban populations with parks that promote well-being.

Publications

Urban Recommendations for Health-promoting Behavioral Change

In this strand of work, we investigate how personalized recommender systems can encourage healthier urban lifestyles through outdoor activities.

We developed a routing engine prototype for running routes in London, leveraging open data from platforms like Flickr, OpenStreetMap, and local crime statistics to characterize the perceived environment. The system recommends two types of routes: scenic paths and safer, more populated alternatives. While most runners preferred scenic routes for their tranquility and aesthetic appeal, safety concerns, especially when running after dark, often shifted preferences toward busier, better-lit streets. These insights demonstrate how context-aware personalization can meaningfully support individual choices.

In another study, we examined how recommender systems could help urban residents discover parks aligned with their leisure and wellness needs. Using survey data and over a million geotagged Flickr images, we found that conventional proximity-based recommendations tended to favor well-known parks, overlooking hidden gems. By contrast, personalized models corrected for this popularity bias, offering more diverse and individually relevant park suggestions.

Publications

Analysis of Point-of-Interest Recommendation

In collaboration with Pablo Sánchez (Universidad Pontificia Comillas, Spain), Alejandro Bellogín (Universidad Autónoma de Madrid, Spain)

Data-Driven Destination Recommender Systems

My doctoral research project at the Technical University of Munich under the supervision of Wolfgang Wörndl and Jörg Ott.

Choosing where to travel is a complex and emotionally significant decision, yet systems to guide destination selection remain underdeveloped compared to those for hotels or restaurants. In our doctoral research, we tackled this gap by focusing on traveler mobility analysis, destination characterization, and conversational recommender systems, aiming to make travel planning more personalized and effective.

To better understand traveler behavior, we analyzed global mobility patterns using data from location-based social networks. Cluster analyses revealed distinct traveler types, showing that ignoring these groups can lead to inaccurate recommendations. We also developed a hierarchical map of travel regions and methods to recommend personalized stay durations, enhancing the relevance of suggestions.

For destination characterization, we evaluated 18 data models, finding that textual sources like Wikipedia effectively captured the essence of “touristic experiences.” By optimizing feature-based models with black-box techniques, we made them competitive with implicit approaches, ensuring they better matched traveler preferences.

Recognizing the challenges users face in articulating preferences and managing expectations, we designed CityRec, a conversational recommender system. Our “Navigation by Revealing Trade-offs” approach helped users clarify their priorities and compromises, making decision-making more informed and satisfying. A study of 600 participants confirmed the system’s effectiveness, demonstrating the value of integrating intuitive design with robust algorithms.

This research advanced destination recommendation by offering new insights into traveler mobility, refining destination characterization techniques, and introducing a user-centered conversational framework to improve travel planning experiences.

Publications