Fall 2020 Student Projects Overview

This article has been written by Ceyhun Burak Akgül and Ahu Sökmenoğlu Sohtorik.

First things first, this has been great. When we started the course last October, we were excited but also curious about two things. First, we didn’t know how many people would enroll in the course. Second and maybe more importantly, how many of them would continue until the end? The answers to both questions turned out to be fabulous… beyond expectations. Initially, we thought we could be useful for 15 people in a graduate course like this, but in the face of the interest we have seen, we had to almost double the capacity. Basically, we didn’t say no to anyone interested and we started with more than 25 graduate students. It is not uncommon for new graduate courses with exciting names like ours to attract many students, but keeping the same audience alive and active throughout the semester is never an easy feat. Well… we managed, we left almost no person behind. We all have passed a wonderful semester in terms of attendance, participation, and the outcome. This post is a brief summary of what we did throughout the semester and what students produced.

Together with my friend and colleague Dr. Ahu Sökmenoğlu, we had constructed the course along two initial axes that would eventually (and hopefully) join into one. On one hand, Ahu, building on her PhD thesis and work thereafter, described the urban problem from many dimensions and paved the way for us to make a case for employing data analytics as a new urban lens. On the other hand, I introduced AI in general, and data analytics in particular, from a technical perspective in a way that can be hopefully easily but also productively grasped. The audience proved us very satisfied. In fact, that’s the loveliest thing about teaching experience, seeing the result where you put the effort in. Although the background of our audience was in general rudimentary when it comes to AI and data analytics; their interest, hard work, and diligence made them succeed through the toughness of the concepts and techniques we presented during the semester. The most difficult part however, and pretty much expectedly, has been to join these two axes into one in order to deliver the promise of the course: How can data analytics be useful for urban problems?

The answer to this question has been where we have put the most effort in through several interactive discussions and iterations. Our foundational teaching aims in this course were five-fold, and they translate into the following imperatives for the student:

ASK a data analytic question properly, starting from a bigger, overarching urban issue.

DESIGN a data model and select attributes that are most relevant to the urban question at hand.

LEARN the elementary dynamics of the standard data analytics toolkit, which consists of descriptive analysis, prediction, clustering, and association learning.

APPLY the fundamental data analytics techniques such as data preparation, tabularization, and transformation; correlation and mutual information analyses; decision trees and the naive Bayes method for classification; linear regression; k-means and hierarchical clustering methods, and association rule mining. As application software, our choice has been the very intuitive and easy-to-use RapidMiner.

INTERPRET the findings and present (visualize) them in a way to trigger further questions going beyond the scope of a one-semester introductory course.

These foundational aims not only determined the way we have delivered the lectures but also served as a guideline for the students to tackle their projects in a systematic way.

The turnout of successful projects which made it to the end of the semester has been quite satisfactory. 22 students out of the 26 enrolled submitted and presented their projects at the finals (complete list can be found in the later sections of this post with links to project reports). We made a categorical distribution of the works based on the keywords submitted by the authors and on our individual assessment on the kind of urban question they addressed. All projects shared many urban concepts and issues in their undertakings and 50% of them employed at least one type of morpho-typological urban attribute. Considering the kind of urban question as a key, we observed that the following urban categories stand out in the projects:

(1) Urban Livability,

(2) Urban Elements,

(3) Urban Transformation,

(4) Urban Transportation,

(5) Socioeconomic,

(6) Sociopolitical.

Urban Livability

As seen in the figure above, a great majority (41%) of the students addressed the issue of Urban Livability. We believe this particular concentration is no coincidence because the quality of life in urban environments is one of the major concerns of our times. Istanbul, for the obvious fact that ourselves and our audience are living mostly in Istanbul, was the most popular city focused on in the projects, and Istanbul suffers a lot from the lack of urban amenities although it’s one of the most admired and prominent urban spots across the world.

In the Urban Livability category, individual projects addressed similar subtopics. For instance, Elif Alkılınç and Nur Sipahioğlu independently addressed the emerging concept of walkability in a city. Again, Istanbul, all its presumed beauty aside, is not a triumph of pedestrian comfort. As such, Elif’s and Nur’s projects went into the heart of one of the major bottlenecks in Istanbul: Walking is hard in Istanbul, that’s not just because of the potential energy residing in its topography (Istanbul is built on seven hills and has grown into many through the ages) but also because streets of Istanbul are either not properly designed for pedestrians or are occupied by vehicles and other obstacles blocking the sidewalks.

Using decision tree, naive Bayes and K-means clustering methods, Elif Alkılınç explored the characteristics of walking routes and drivers of route ratings by analyzing 68 walking routes, 34 of them borrowed from the “Istanbul by Walking” by Uslu (2018).

Another common topic was public stairs, which are an indispensable part of the citizens’ everyday in Istanbul. They are a natural part of Istanbul’s challenging yet rewarding topography providing magnificent city views. While Hüseyin Kadıoğlu analyzed Istanbul streets with public stairs from an aesthetics perspective, Sheida Shakeri focused on a more social aspect and explored the potential of public stairs of Istanbul in gathering people together. Sheida further elaborates on Grafitti In Istanbul stairs in her beautiful Medium article.

Sheida Shakeri focused on the public space formed around the city stairs and using decision tree and naive Bayes classification, she explored the effects of the above visualized attributes in encouraging the art of graffiti.

Paucity of green spaces is also a major issue in Istanbul. There are not so many of them for public use and the ones that are already available are not as much used for recreation as in a typical European city. This is in stark contrast to the fact that Istanbul is naturally green; it not only hosts scattered lines of trees especially along its Bosphorus strait (and actually pretty much everywhere) but it’s also surrounded by urban forests (which are unfortunately endangered due to huge construction projects). Sarvin Eshaghi asked whether data mining can be leveraged to understand why existing urban green spaces in Istanbul are seldom used.

Confirming the well-known statement “Shopping is arguably the last remaining form of public activity” expressed by Koolhas and his colleagues, shopping has become the main urban activity in Turkey since 20 years or so. While reasons behind this socio-spatial destruction of urban life are beyond our context, it’s perfectly legitimate to ask whether the availability and the design quality of urban equipment affects the popularity of shopping streets. Deniz Tutucu looked into this question from a data analytics perspective.

More holistic questions have been asked in this category. Dilan Öner explored the relationships between morphological urban characteristics, availability of cars/people/trees in the streets, environmental and climatic parameters, and the way citizens perceive them in terms of comfort and satisfaction. Dilan’s prospective data analytics project programme is very rich and interesting in its ambitions to leverage visual, auditory and sensorial data to address the measurability of urban comfort.

The picture above illustrates the relationship between subjective evaluation of the pedestrians in terms of environmental and climatic comfort and other urban characteristics, explored by Dilan Öner using the Random Forest algorithm.

Selen Aksoy, starting from Lefebvre’s trilogy of the designed, perceived, and lived space, questioned which urban attributes contribute to the “publicness” of public spaces and leveraged data analytics to provide exploratory answers.

The picture above depicts attribute classification in Selen Aksoy’s work; she divided her study area (Beşiktaş) into pixels and gathered data for each pixel using OpenStreetMap, ArcGIS and Flickr.

The last work in the Urban Livability category addressed one of the grim aspects of living in cities. Barış Terzi tried to answer whether certain street characteristics lead to increased crime rates, and if yes, which type of crimes go together with which type of urban characteristics.


Urban Elements

We consider this category as a salutation to Kevin Lynch’s seminal book, The Image of The City. Although we didn’t receive enough projects to cover all of the original Lynchian elements (paths, edges, districts, nodes, and landmarks) that make up the image of a city according to Lynch, we were happy to see the following three works on Urban Elements.

Eda Özgener, inspired by her personal observation on the resemblance between Karaköy port and Lucerne Seabrück area (Switzerland), has brought one of the most attractive urban features in Istanbul into the spotlight and she decided to use seashore segments as her working urban unit (which can be seen as anyone of the Lynchian elements depending on the city and depending on the specific location). Her findings exploratorily confirmed the premised resemblance. While generalization of data analytics models requires much more data in amount and diversity, this alignment between inspiration and finding is promising and motivating by itself.

Two other projects both addressed the urban square, which can be a district, a node or a landmark following the Lynchian terminology. Nisa Çelebi’s project was very interesting in the type of attributes she used as input to a predictive process. She employed sound characteristics of Istanbul squares in order to determine their primary category: an urban icon (landmark), ceremony/rally place (district), transportation hub (node) or cultural/commercial center (district).

For future work, Nisa Çelebi envisions an online interactive interface (image above) integrated with a real-time sound gathering system instantaneously mining sound data to explore the sound identity of city squares.

Nurdan Akman dealt with the more holistic question of defining city squares from their urban characteristics such as function and activities they support, their popularity, their landmark aspects, their accessibility, their surface area, their geometry and their popularity.


Urban Transformation

This category is a perfect example of the blessings we had in terms of projects. The two works we categorized as Urban Transformation were very original in their subject matter and in their approach to data analytics.

Dilara Demiralp introduced the concept of magnetic giants as “big, uncommunicative, impermeable, out of scale, unfriendly, inaccessible” urban units such as stadiums. Considering the homes to two of the three most prominent football clubs in Istanbul as magnetic giants, she explored the potential of much tinier surrounding units such as co-working spaces, restaurants and cafes, retail stores transforming into something else under the supposed magnetism of these mostly dormant but biweekly roaring urban giants. Interestingly, the pandemic temporarily transformed these giants themselves into grotesque urban non-necessities: Under the pandemic, a stadium and a modest cafe around it are no longer distinguishable in terms of their social effects. Still, the underlying data analytic question is very interesting and should be revisited in a pandemic-free future.

Dilara Demiralp explored the effects of the “magnetic giants” of Istanbul over other urban functionalities using data mining methods and she envisions that her work can be further developed by including the magnetic potentials of other lower-scale functionalities as sketched above.

Selen Çiçek’s project is somewhat looking backwards in the history of a city and questions how we can detect and repurpose decaying urban units. Selen’s data analytic playground has been Izmir’s, the third biggest city in Turkey, now decaying neighborhoods hosting old landmark buildings, which are no longer in use. She contrasts these neighborhoods with “ideal” ones from Vienna, which is an old but one of the most livable cities across the world. By contrasting and comparing the urban data attributes of two sets of old neighborhoods, one set from the decaying Izmir and the other from the thriving Vienna, Selen hoped to learn from Vienna and find indications on how to construct and implement regenerative urban policies for Izmir, which would prevent urban decay.


Urban Transportation

We would be surprised if Urban Transportation didn’t emerge as a category. Today, commuting is probably the most important pain point in urban space, with drastic negative effects on the environment, on the economy, and all in all, on overall citizen happiness.

As a partial remedy to the urban mobility problems, which have increased enormously in Istanbul, a city with a population of over 15 million residents, Istanbul Metropolitan Municipality introduced ISBIKE — a public bicycle sharing system. Ali Yılmaz quantified and visualized the activity level of municipal bike stations (isbike) across Istanbul, and tried to relate these activity levels with other urban attributes surrounding the stations. He employed real-time and publicly available usage data provided by Istanbul Metropolitan Municipality. We consider his work as a promising attempt to provide citizen insights to city administrators. Istanbul, not a pedestrian-friendly city, is not a biker-friendly city either. Such quantifications and visualizations can potentially expose these aspects as has never done before.

Comparing data mining analysis results with above illustrated GIS based visualizations, Ali Yılmaz’s work raised new research questions regarding the relationships of the activity levels of ISBIKE stations with the surrounding urban functionalities.

Ekin Kılıç’s project addressed a sort of an urban chicken-and-egg problem. Should we build more motorized vehicle roads, metro lines, or railway to accommodate more people in a city or does the very fact of building more transportation networks lead to more people in a city? Data analytics constitutes a perfect means to answer such questions provided there is the right data in place in kind and amount. By modeling the predictive relationship from the size of the rail network to the population, Ekin formulated the right way to look at the chicken-and-egg problem. Although his findings have been inconclusive for the time being, they mediated the emergence of new research questions.

Erenalp Saltık’s project investigated the weird interplay between public transportation availability and rental value. In a hypothesized world where convenience means more value, one would expect a direct relationship. Alas, it’s not been the case for Istanbul (and probably for elsewhere in the world), data show that places which are more accessible by public transportation are not necessarily the most valuable ones. Sometimes, screening factors are more influential than the available data.



We had three works in the expectedly heterogeneous Socioeconomic category. Ece Buldan’s work went beyond borders and investigated how immigrants from the UK to southern Turkey feel themselves: Are they now part of where they moved to or do they still feel as “others” to their new environment? And which factors are determinant in this interesting dichotomy?

Emine Zeytin’s work addressed the question of urban diversity in the UK boroughs and used hate crime rates as an indicator of the associated mishaps. While diversity is conducive to more interestingness and creativity, it can also have adverse effects. As such, the study addresses a very intricate domain where data analytics should be very carefully employed.

The diagram above illustrated by Emine Zeytin conceptually depicts the relationship of urban diversity with hate crime.

Tahsin Oğuz Koç produced a set of exploratory results using a set of 10 urban and socioeconomic attributes of Istanbul districts. His findings establish that the human development index of citizens and regional development index of the places they live are correlated and carry information on each other, waiting for further investigation.



We tagged the last two works in our overview as Sociopolitical. Interestingly two students, Feyza Çınar and Gizem Fidan, asked more or less the same perfectly legitimate data analytic question: What makes an urban environment more favorable for protests and demonstration? In the context of our beloved but strange country, Turkey, this legitimate question is also unnecessarily controversial due to authoritarian intolerance. Leaving this controversy aside, both of the works were very interesting. Feyza explored the relationship from three different sets of urban attributes of different nature to the occurrence of demonstration in designated Ankara spots: (1) attributes coding the morphological characteristics, (2) attributes coding the environmental conditions, and (3) attributes coding transportation convenience. Gizem on the other hand, focused on the urban design competitions organized by Istanbul Metropolitan Municipality for historic city squares of Istanbul; Taksim, Bakırköy and Kadıköy. She was inspired by the criticism towards the winning projects for not considering the permeability of the squares for protests. Starting from a similar set of attributes, analyzed new design proposals for public spaces in Istanbul in terms of their suitability for protests. This particularly creative usage of data analytics in the urban context is quite interesting in that it can potentially reveal the unintended biases of urban designers.

The image above, prepared based on the results of the K-means analysis implemented by Gizem Fidan, geographically maps the clustering of protest spaces in Istanbul and also informs about the most effective variables in the process of clustering.


To conclude (for the time being)…

Tough questions stand.

As designers, do we design to please citizens or the authority?

As citizens, do we have a say on where we are supposed to live?

Ahu & Ceyhun, March 2021.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store